carlos.perezuv. es
reporting summary for nature research说明书
|April2018Corresponding author(s):Reporting SummaryNature Research wishes to improve the reproducibility of the work that we publish.This form provides structure for consistency and transparency in reporting.For further information on Nature Research policies,see Authors &Referees and the Editorial Policy Checklist .Statistical parametersWhen statistical analyses are reported,confirm that the following items are present in the relevant location (e.g.figure legend,table legend,main text,or Methods section).The exact sample size (n )for each experimental group/condition,given as a discrete number and unit of measurementAn indication of whether measurements were taken from distinct samples or whether the same sample was measured repeatedlyThe statistical test(s)used AND whether they are one-or two-sidedOnly common tests should be described solely by name; describe more complex techniques in the Methods section.A description of all covariates testedA description of any assumptions or corrections,such as tests of normality and adjustment for multiple comparisonsA full description of the statistics including central tendency (e.g.means)or other basic estimates (e.g.regression coefficient)ANDvariation (e.g.standard deviation)or associated estimates of uncertainty (e.g.confidence intervals)For null hypothesis testing,the test statistic (e.g.F ,t ,r )with confidence intervals,effect sizes,degrees of freedom and P value notedGive P values as exact values whenever suitable.For Bayesian analysis,information on the choice of priors and Markov chain Monte Carlo settingsFor hierarchical and complex designs,identification of the appropriate level for tests and full reporting of outcomesEstimates of effect sizes (e.g.Cohen's d ,Pearson's r ),indicating how they were calculatedClearly defined error barsState explicitly what error bars represent (e.g. SD, SE, CI)Our web collection on statistics for biologists may be useful.Software and codePolicy information about availability of computer codeData collection Data analysis For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature,software must be made available to editors/reviewers upon request.We strongly encourage code deposition in a community repository (e.g.GitHub).See the Nature Research guidelines for submitting code &software for further information.Keiji KubaBio-Rad Image lab Touch software (version 2.1.0.35),Thermal Cycler Dice Real Time System Software (version 5.11C for TP900),BDCellQuest ProTM (version 6.0),Nikon NIS-Elements (version 3.22.00),Bio-Rad Microplate Manager Software (version 6.3),FinePoint(version 2.8.0.12146),FACS Calibur (Becton Dickinson)Bio-Rad Image lab (version 5.2),Thermal Cycler Dice Real Time System Software (version 5.11C for TP900),Excel (version 2108),Prism(version 9.2.0),BD CellQuest ProTM (version 6.0),FlowJo (version 10.8.0),FinePoint (version 2.8.0.12146)April 2018DataPolicy information about availability of dataAll manuscripts must include a data availability statement .This statement should provide the following information,where applicable:-Accession codes,unique identifiers,or web links for publicly available datasets-A list of figures that have associated raw data-A description of any restrictions on data availabilityField-specific reportingPlease select the best fit for your research.If you are not sure,read the appropriate sections before making your selection.Lifesciences Behavioural &socialsciences Ecological,evolutionary &environmental sciencesFor a reference copy of the document with all sections,see /authors/policies/ReportingSummary-flat.pdfLife sciences study designAll studies must disclose on these points even when the disclosure is negative.Sample size Data exclusions Replication Randomization Blinding Reporting for specific materials,systems and methodsUnique biological materialsPolicy information about availability of materialsObtaining unique materials AntibodiesAntibodies used The data that support the findings of this study are provided in the Article and its Supplementary Information.Source data are provided with this paper.GenBank accession code for SARSCoV-2,HKU-001a is MT230904.1.We performed the power analysis to predetermine sample size.We did not exclude the data.One independent experiment was performed for Figures 1b,2b-d,5b,5c and Extended Data Figures 1a-h,3c,3d,6b,6f.Independentexperiments were performed two times for Figures 1c,2f-j,3c-l,4b-k,5a,5d-h and Extended Data Figures 1i,1j,3a,3b,5a-f,6a,6c-e,7a-g,8a-f and three times for Figures 6b-g and Extended Data Figures 2a-e,4a-h,and consistent results were obtained.The animals were allocated into experimental group in random.The investigators were blinded to group allocation during data collection and analysis.We confirm that the unique materials in this study are available from us.For western blotting,we used anti-mouse ACE2antibody (Crackower,M.et al.,Nature,2002,doi:10.1038/nature00786),anti-hamster GAPDH antibody (GeneTex,GTX100118),anti-SARS-CoV/SARS-CoV-2NP antibody (Chan,J.F.et al.,Clin Infect Dis,2020,doi:10.1093/cid/ciaa325),anti-human ACE2antibody (R&D systems,MAB9331),anti-beta-actin antibody (Sigma,A5316,batchApril 2018Validation Eukaryotic cell linesPolicy information about cell linesCell line source(s)Authentication Mycoplasma contamination Commonly misidentified lines(See ICLAC register)Animals and other organismsPolicy information about studies involving animals ;ARRIVE guidelines recommended for reporting animal researchLaboratory animals Wild animals Field-collected samples Flow CytometryPlotsConfirm that:axislabels state the marker and fluorochrome used (e.g.CD4-FITC).axis scales are clearly visible.Include numbers alongaxes only for bottom left plot of group (a 'group'is an analysis of identical markers).plots are contour plots with outliers or pseudocolor plots.numerical value for number of cells or percentage (with statistics)is provided.MethodologySample preparation Instrument Software Cell population abundance number 123M4876)and anti-human IgG antibody (MBL,103R,lot186).For in vitro binding assay,we used anti-human ACE2antibody (Novus Biological,SN0754,NBP2-67692,lotHN0420),anti-mouseACE2antibody (Crackower,M.et al.,Nature 2002,doi:10.1038/nature00786),anti-human IgG antibody (MBL,103R,lot186),anti-B38-CAP polyclonal antibody (Minato,et al.,Nat Commun,2020,doi:10.1038/s41467-020-14867-z),FITC-conjugated humanIgG-specific polyclonal antibody (Jackson ImmunoResearch,#109-095-088,lot137124)and Fc antibody (JacksonImmunoResearch,#109-035-098,lot146365).Validation information of each antibody is as follows:anti-mouse ACE2antibody (Crackower,M.et al.,Nature 2002,doi:10.1038/nature00786),anti-hamster GAPDH antibody (https:///PDF/Download?catno=GTX100118),anti-SARS-CoV/SARS-CoV-2NP antibody (Chan,J.F.et al.,Clin Infect Dis,2020,doi:10.1093/cid/ciaa325),anti-human ACE2antibody (https:///pdfs/datasheets/mab9331.pdf?v=20211006&_ga=2.210780560.420495370.1633586814-1097249821.1598599156),anti-beta-actin antibody (https:///deepweb/assets/sigmaaldrich/product/documents/863/388/a5316blot.pdf),anti-human ACE2antibody(https:///PDFs4/NBP2-67692.pdf),anti-human IgG antibody (https://ruo.mbl.co.jp/bio/dtl/A/?pcd=103R),anti-B38-CAP polyclonal antibody (Minato,et al.,Nat Commun,2020,doi:10.1038/s41467-020-14867-z),FITC-conjugatedhuman IgG-specific polyclonal antibody (https:///catalog/products/109-095-088)and Fc antibody(https:///catalog/products/109-035-098).Vero E6cells (CRL-1586)and Caco2cells (HTB-37)were obtained from ATCC.Vero E6/TMPRSS2(JCRB1819)were obtainedfrom JCRB Cell Bank.Expi293F cells were obtained from Thermo Fisher Scientific (A14635).Vero E6cells and Caco2cells were authenticated with STR profiling by ATCC.VeroE6/TMPRSS2cells was authenticated byJCRB Cell Bank,but no information of technique to authenticate is available.Expi293F cells was not authenticated.Not tested for Mycoplasma contaminationNoneAll animal experiments conformed to the Guide for the Care and Use of Laboratory Animals,Eighth Edition,updated by the USNational Research Council Committee in 2011,and approvals of the experiments were granted by the ethics review board ofAkita University,NIBIOHN,the University of Tokyo or the University of Hong Kong.We used 3,6or 10week-old male or female Syrian hamsters and 3-4month-old male human-ACE2transgenic mice.Three tofour month-old male or female C57BL/6J mice were used to backcross human-ACE2transgenic mice.No wild animals were used in this study.No field collected samples were used in this study.FACS Calibur (Becton Dickinson)FlowJo v10.8softwareFive thousand cells per analysis and 100%purity of live Vero E6cells。
如何利用马尔可夫链蒙特卡洛进行贝叶斯优化(八)
贝叶斯优化是一种用于求解复杂优化问题的方法，它使用贝叶斯推断和概率模型来寻找最优解。
在实际应用中，我们常常需要利用马尔可夫链蒙特卡洛（MCMC）算法来进行贝叶斯优化，下面我们将详细介绍如何利用马尔可夫链蒙特卡洛进行贝叶斯优化。
首先，我们需要了解一下马尔可夫链蒙特卡洛算法的基本原理。
马尔可夫链蒙特卡洛算法是一种通过随机抽样来近似求解概率分布或期望值的方法。
它利用马尔可夫链的性质，通过迭代的方式生成服从目标分布的样本。
在贝叶斯优化中，我们通常需要对目标函数进行采样，然后利用这些样本来逼近目标函数的分布，马尔可夫链蒙特卡洛算法正是可以帮助我们完成这个任务。
其次，我们需要确定目标函数的先验分布。
在贝叶斯优化中，我们通常假设目标函数服从一定的先验分布，然后利用观测到的样本数据来更新这个分布。
确定目标函数的先验分布是贝叶斯优化的关键一步，它直接影响到最终的优化结果。
通常情况下，我们会选择一些常见的先验分布，比如高斯分布、指数分布等，根据具体的问题来确定先验分布。
接着，我们需要选择合适的马尔可夫链蒙特卡洛算法。
马尔可夫链蒙特卡洛算法有很多种，比如Metropolis-Hastings算法、Gibbs抽样算法等，每种算法都有其适用的场景和特点。
在贝叶斯优化中，我们需要根据目标函数的性质和先验分布的特点来选择合适的算法。
通常情况下，Metropolis-Hastings算法是一个比较通用的选择，它适用于各种类型的目标函数和先验分布。
然后，我们需要进行马尔可夫链蒙特卡洛的参数调优。
马尔可夫链蒙特卡洛算法有一些参数需要调整，比如迭代次数、步长大小等。
这些参数的选择对最终的优化结果有很大的影响，因此需要仔细调优。
一般来说，我们可以通过一些自适应的方法，比如随机漫步Metropolis算法、哈密顿蒙特卡洛算法等，来自动调整这些参数，以达到最佳的采样效果。
最后，我们需要进行收敛性和有效性的检验。
马尔可夫链蒙特卡洛算法在实际应用中往往需要一定的迭代次数才能达到收敛，因此需要进行收敛性的检验，以确保采样结果的准确性。
Results of a Study using the Motivation Strategies for Learning Questionnaire (MSLQ) in an Introduct
Results of a Study using the Motivation Strategies for Learning Questionnaire (MSLQ) in an Introductory Engineering Graphics CourseAaron C. Clark1 Jeremy V. Ernst2 Alice Y. Scales3Abstract – This paper will present data related to a study conducted at NC State University in the spring of 2008 that focused on student motivation in an introductory graphics course. This study conducted a motivation and learning assessment using the Motivated Strategies for Learning Questionnaire (MSLQ) Attitude Survey. The motivational portion of MSLQ focuses on six areas associated with student learning and motivation. These areas were intrinsic goal orientation, extrinsic goal orientation, task value, control of learning beliefs, self-efficacy learning performance, and test anxiety. Findings from the study included the identification of enduring motivational factors for learning graphics education. Insights into the strategic learning process of students in a graphics education course will be discussed. Also, areas of concern for future pedagogical development and course improvement will be highlighted.Keywords: MSLQ, Introductory Graphics Course,I NTRODUCTIONMany motivational processes are responsive to individual properties associated with tasks, the classroom, or the context within student engagement [Wolters & Pintrich, 11]. Literature on student motivation identifies many beliefs and constructs, but control, competence, and self-regulated strategic learning remain chief among them [Shell & Husman, 9]. Internal pressures also serve as strong motivators in adult learners [Knowles, Holton, & Swanson, 4, pp. 64-66]. An attitude of self-determination resides at the nucleus of intrinsic motivation [Johari & Bradshaw, 5]. This self-determined attitude is primarily a result of feeling competent and/or independent. In adults, feelings of intellectual competence can be highly motivational when paired with internal pressures that serve as a driving force. Self-determination theory research has placed a large amount of attention on, not only intrinsic motivation, but also extrinsic motivation. Extrinsic motivation refers to “engaging in an activity to obtain an outcome separable from the activity itself” [Vansteenkiste, Timmermans, Lens, Soenens, & Van den Broeck, 10, pp. 388]. A study conducted by Bye, Pushkar, & Conway [2] at Concordia University identifies intrinsic motivation as a predictor of positive classroom effect, while self-improvement and personal growth were found to be highly valued in comparison with extrinsic goals, further distinguishing between intrinsic and extrinsic motivation.1 NC State University, Box 7801, Raleigh, NC 27695-7801, aaron_clark@2 NC State University, Box 7801, Raleigh, NC 27695-7801, jeremy_ernst@3 NC State University, Box 7801, Raleigh, NC 27695-7801, alice_scales@Student motivation possesses a value component involving students’ goals and beliefs about the importance of a task or their personal interest in an application. Motivational value has been conceptualized through various approaches (e.g., learning vs. performance goals, intrinsic vs. extrinsic orientation, task value, and intrinsic interest); this motivational component effectively concerns students' motives for the completion of a task [Pintrich & De Groot, 8]. Beyond beliefs pertaining to importance and interest is self-efficacy. Students’ perceived self-efficacy might influence the process by which he or she selects activities to participate in or complete. There are many circumstances where students assume and perform activities they deem themselves capable of successfully completing and avoid those they believe exceed their ability [Yang, 12]. This paper will examine the results of a study conducted at North Carolina State University that looked at the type of motivation exhibited by students taking an introductory engineering class.M OTIVATED S TRATEGIES FOR L EARNING Q UESTIONNAIRET he Motivated Strategies for Learning Questionnaire (MSLQ) is an instrument designed to evaluate “college students’ motivational orientation and use of different learning strategies for a college course” [Pintrich, Smith, Garcia, and McKeachie, 8]. The broad cognitive analysis of motivation and learning strategy, paired with the social cognitive view of motivation and self-regulated learning, serves as the foundation of MSLQ. The MSLQ consists of two major sections: a motivation section and a learning strategies section. The motivation segment has 31 items that evaluate students’ goals and value beliefs, students’ beliefs about skills necessary to succeed, and test anxiety associated with a specific course [Duncan & McKeachie, 3]. Duncan & McKeachie further differentiate the learning strategy section of the MSLQ as identifying students’ use of different cognitiv e and metacognitive strategies as well as student management of resources. The motivation section and the learning strategies section of the MSLQ include 81 items. Each item is rated using a 7-point Likert-type scale. The rating scale ranges from one (not at all true of me) to seven (very true of me).Pintrich, Smith, Garcia, & McKeachie [8] describe the motivation scales of the MSLQ as vehicles to acquire information associated with value, expectancy, and affect. Value assists in exploring intrinsic and extrinsic goal orientation, expectancy targets beliefs about learning and self-efficacy, and affect gauges test anxiety. Learning strategies investigated through the motivation scales are drawn from a broad compilation of cognitive research representing cognitive processing and its affect on student learning [Lynch, 6].Numerous MSLQ studies have been conducted that present evidence of internal consistency, reliability, and predictive validity of the instrument [Pintrich, Smith, Garcia, & McKeachie, 8; Artino, 1; Duncan & McKeachie, 3]. The MSLQ represents a method to accurately and holistically gage student motivation and self-regulated learning grounded by a theoretical basis. The MSLQ allows student learning researchers to move beyond traditional examinations of individual differences in learning styles to gain insight into the motivation and learning specifically occurring in a targeted college course. In this investigation, an introductory engineering graphics course wasselected to investigate intrinsic goal orientation, extrinsic goal orientation, task value, control of learning beliefs, self-efficacy learning performance, and test anxiety with the MSLQ Attitude Survey.M ETHODOLOGYThis targeted investigation utilized the results of 31 motivational questions MSLQ Attitude Survey to examine six proposed null hypotheses concerning motivation and satisfaction of student learning. These null hypotheses were: 1) Ho: Student intrinsic goal orientation elements are independent components of motivation and learning. 2) Ho: Student extrinsic goal orientation elements are independent components of motivation and learning. 3) Ho: Student task value elements are independent components of motivation and learning. 4) Ho: Student controls of learning beliefs are independent components of motivation and learning. 5) Ho: Student self-efficacy and learning performance elements are independent components of motivation and learning. 6) Ho: Student test anxiety elements are independent components of motivation and learning.These hypotheses guided the motivation and learning investigation utilizing the MSLQ Attitude Survey as the means for data acquisition. Specifically, the six hypotheses structure the investigation to identify enduring motivational factors for learning graphics in the introductory engineering graphics course at NC State University.To better gauge indicators of student attitude and motivation, the MSLQ data analysis was shortened. As prescribed by Matthews [7] to solely measure motivation concerning goal orientation, extrinsic goal orientation, task value, control of learning beliefs, self-efficacy learning performance, and test anxiety, the MSLQ analysis was limited to 31 questions specifically targeted to student motivation. Additionally, Matthews identified the MSLQ item equivalent subsets to provide a targeted analysis of the six focal areas associated with student learning and motivation.In the 10th week of the 2008 spring semester the course instructors administered the MSLQ instrument to student participants in the introductory engineering graphics course. The questionnaire took the participants approximately 15 minutes to complete. One hundred and sixty one students in seven separate sections of GC 120 (Foundations of Graphics) completed and returned the instrument. One of the 161 participants failed to complete items 24 and 29 of the targeted subgroup analysis, but the researchers decided to include this questionnaire in the completed group. The researchers gathered the completed instruments from the course instructors, entered the MSLQ data, tabulated the questionnaire results, analyzed the target items, and formed conclusions based on the six identified student learning and motivation areas.R ESULTSThe proposed hypotheses were evaluated using a one-sample calculation of variance. The test of independence tabulates MSLQ instrument items in their designated categories and computes a chi-square value. This procedure uses the critical value to evaluate the proportional value derived from the Chi-Square table. A significant p-value foran item in a category demonstrates that it is independent of the other items and, therefore, has no relationship to the other items in its category or the category itself.The identified MSLQ item equivalents to investigate intrinsic goal orientation were 1, 16, 22, and 24 (See Table 1). Within the item equivalents that measured intrinsic goal orientation, item 16 had the highest average, while item 24 had the lowest. As a group, the intrinsic goal orientation items averaged 4.68 on the seven-point scale. The sampling variance, reported in the data summations, was due to a statistical fluctuation in the responses on intrinsic goal orientation sub grouped items identified in the six student learning and motivation areas. Additionally, evaluation of the chi-square statistic and the proportional value associated with each item identified all four MSLQ items within their student learning and motivation area as significantly different from one another, given the predetermined alpha level of significance (0.05). Items 1, 16, 22, and 24 all had p-values smaller than 0.05, therefore the null hypothesis that intrinsic goal orientation elements are independent components of motivation and learning could not be rejected because there is evidence that the questions were independent of the category and each other by virtue of their significant p-values.Table 1. MSLQ Intrinsic Goal OrientationThe identified item equivalents to investigate extrinsic goal orientation were MSLQ items 7, 11, 13, and 30 (See Table 2). Within the item equivalents of extrinsic goal orientation, item 13 had the highest average, while item 30 had the lowest. As a group, the extrinsic goal orientation items averaged 5.35 on the seven-point scale. Additionally, reporting and evaluation of the chi-square statistic and the proportional value associated with each item identified three of the four items were significantly different from one another. Item 13 was found not to significantly differ within the subgroup. Items 7, 11, and 30 all had a p-value smaller than 0.05, therefore, the null hypothesis that statedthat extrinsic goal orientation elements are independent components of motivation and learning also failed to be rejected.Table 2. MSLQ Extrinsic Goal OrientationThe identified item equivalents to investigate task value were MSLQ items 4, 10, 17, 23, 26, and 27 (See Table 3). Within the item equivalents for task value, the six items provide participant averages relatively close to one another. As a group, the task value items averaged a 5.16 on the seven-point scale. The sampling variance again was due to a statistical fluctuation in participant responses on the task value sub grouped items. Likewise, reporting and evaluation of the chi-square statistic and the proportional value associated with each item identified all six of the MSLQ items within their student learning and motivation area as significantly different from each other. The p-values for items 4, 10, 17, 23, 26, and 27 were all lower than the established cut-off value of 0.05, therefore, the null hypothesis that stated that task value elements are independent components of motivation and learning could not be rejected.Table 3. MSLQ Task ValueThe identified item equivalents that examined control of learning beliefs were MSLQ items 2, 9, 18, and 25 (See Table 4). Within the item equivalents of control of learning beliefs, item 18 had the highest average while item 25 had the lowest. As a group, the control of learning beliefs items averaged 5.62. The sampling variance was due to the variation in the participants’ responses on control of learning beliefs sub grouped items identified within the six student learning and motivation areas. The reporting and evaluation of the chi-square statistic, and the proportional value associated with each item, identified three of the four MSLQ items within their student learning and motivation area as significantly different from one another, given the predetermined alpha level of significance (0.05). Item 18 was found not to differ within the response subgroup. Items 2, 9, and 25 had a p-value lower than the critical value of 0.05, therefore, again the results failed to reject the null hypothesis that control of learning beliefs is an independent component of motivation and learning.Table 4. MSLQ Control of Learning BeliefsThe identified item equivalents to investigate self-efficacy learning performance are MSLQ items 5, 6, 12, 15, 20, 21, 29 and 31 (See Table 5). Within the item equivalents of self-efficacy learning performance, the eight items present participant averages relatively close to one another. As a group, the self-efficacy learning performance items averaged a 5.47 on a seven-point scale. The sampling variance again is due to the statistical fluctuation in participant response on this sub group of items. Additionally, the evaluation of the chi-square statistic and the proportional value associated with each item identified six of the eight MSLQ items within their student learning and motivation area as significantly differing from one another based on the predetermined alpha level of significance (0.05). Items 20 and 21 were found not to significantly differ within the response subgroup; however, items 5, 6, 12, 15, 29 and 31 were lower than the critical p-value set at 0.05; therefore, it was not possible to reject the null hypothesis that self-efficacy and learning performance are independent components of motivation and learning.Table 5. MSLQ Self-Efficacy Learning PerformanceThe identified item equivalents to investigate test anxiety are MSLQ items 3, 8, 14, 19, and 28 (See Table 6). Within the items used to examine test anxiety, item 14 had the highest average while item 3 had the lowest. As a group, the task value items averaged 3.74 on the seven-point scale. The sampling variance was again due to the fluctuation in participants’ responses. Evaluation of t he chi-square statistic and the proportional value associated with each item indicated that all five of the MSLQ items significantly differed from each other and were smaller than the predetermined value for significance. Since items 3, 8, 14, 19, and 28 were not found to be significant, the null hypothesis that test anxiety is an independent component of motivation and learning failed to be rejected.Table 6. MSLQ Test AnxietyC ONCLUSIONSItem 13 (“If I can, I want to get better grades in this class than most of the other students”); in the Extrinsic Goal Orientation subgroup, item 18 (“If I try hard enough, then I will understand the course materials”); in the Control of Learning Beliefs subgroup, item 20 (“I’m confident I can do an excellent job on the assignments and test in this course”) and item 21 (“I expect to do well in this class”) of the Self-Efficacy Learning Performance subgroup were identified by the study as continuing motivational and learning factors for learning engineering graphics in the introductory engineering graphics course at NC State University. Considering the fact that these statements “standout” among the others and that each in some way is associated with the level of understanding and the grade they wish to receive in class, grades are still a good motivation factor to consider with these participants. The ability to do well and see relevance in what is being taugh t is also paramount to a student’s motivation in a course, like a fundamentals of engineering graphics. From the data collected for this study, it can be observed that grades, relevance of content, and understanding subject matter are the main factors tha t affect students’ motivation. Based on these findings, more research in areas of strategic learning of students in engineering graphics courses as it relates to their abilities to be self-motivated needs to be conducted, particularly as the structure and delivery methods of engineering graphics courses are rapidly changing. Also, considering the change and growth of new areas and concepts in the engineering graphics profession, how can we utilize contemporary methods to increase student motivation? Again, more investigation is needed in this area of student motivation as the profession works to educate future professionals that use graphics for the 21st century.R EFERENCES[1] Artino, A.R. (2005). Review of the Motivated Strategies for Learning Questionnaire, ERIC documentsED499083.[2] Bye, D., Pushkar, D. & Conway, M. (2007). Motivation, interest and positive affect in traditional andnontraditional undergraduates. Adult Education Quarterly, 60, # 9, PP1275-1288.[3] Duncan, T.G. & McKeachie, W.J. (2005). The making of the Motivated Strategies for Learning Questionnaire.Educational Psychologist. 40(2), 117-128.[4] Knowles, M., Holton, E., & Swanson, R. (1998). The adult learner: The definitive classic in adult education andhuman resource development. Burlington, MA: Gulf Professional Publishing.[5] Johari, A. & Bradshaw, A.C. (2006). Project-based learning in an internship program: A qualitative study ofrelated roles and their motivational attributes. ETR&D.[6] Lynch, D.J. (2006). Motivational factors, learning strategies and resource management as predictors of coursegrades. College Student Journal.40(2), 423-428.[7] Matthews, B. (2004). The effects of direct and problem-based learning instruction in an undergraduateintroductory engineering graphics course. Unpublished doctoral dissertation, North Carolina State University, Raleigh, NC.[8] Pintrich, P.R. (1999). The role of motivation in promoting and sustaining self-regulated learning. InternationalJournal of Educational Research. 31(6), 459-470.[9] Shell, D. F., Husman, J. (May, 2008). Control, motivation, affect, and strategic self-regulation in the collegeclassroom: A multidimensional phenomenon. Journal of Educational Psychology. Vol 100(2), 443-459.[10] Vansteenkiste, M., Timmermans, T., Lens, W., Soenens, B., Van den Broeck, A. (May, 2008). Does extrinsicgoal framing enhance extrinsic goal-oriented individuals' learning and performance? An experimental test of the match perspective versus self-determination theory. Journal of Educational Psychology. Vol 100(2), 387-397.[11] Wolters, C.A. & Pintrich, P.R. (1999). Contextual differences in student motivation and self-regulated learningin mathematics, English, and social studies classrooms. Instructional Science, 26: 27-47.[12] Yang, N.D. (1999). The relationship between EFL learners' beliefs and learning strategy use. System. 27(4), 515-535.Aaron C. ClarkAaron C. Clark is an Associate Professor of Graphic Communications and Technology Education at North Carolina State University in Raleigh, North Carolina. He received his B.S. and M.S. in Technology and Technology Education and earned his doctoral degree in Technology Education. His teaching specialties are in visual theory, 3-D modeling, gaming, and technical animation. Research areas include graphics education, leadership, andscientific/technical visualization. He presents and publishes in both technical/technology education and engineering education. He is currently a Co-PI on grants related to visualization and education and has started new research in areas related to STEM integration and gaming.Jeremy V. ErnstJeremy V. Ernst is an Assistant Professor in the Department of Mathematics, Science, and Technology Education at North Carolina State University. He currently teaches a variety of courses and supervises student teachers in the Technology Education Program. Jeremy specializes in research involving instruction, learning, and visualization for university students, students with disabilities and other at-risk populations in Career and Technical Education. He also has curriculum research and development experiences in technology, trade and industrial education.Alice Y. ScalesAlice Y. Scales is an Assistant Professor and the Assistant Department Head of the Department of Mathematics, Science, and Technology Education at North Carolina State University. She has taught at NC State University since 1988. She has a B.S. in Science Education, a M.Ed. in Industrial Arts Education, and an Ed.D. in Occupational Education. She currently teaches courses in desktop publishing, website development, and introductory engineering graphics.2009 ASEE Southeast Section Conference。
ooDACEToolboxAFlexibleObject-OrientedKriging…
Journal of Machine Learning Research15(2014)3183-3186Submitted6/12;Revised6/13;Published10/14ooDACE Toolbox:A Flexible Object-Oriented Kriging ImplementationIvo Couckuyt∗********************* Tom Dhaene******************* Piet Demeester*********************** Ghent University-iMindsDepartment of Information Technology(INTEC)Gaston Crommenlaan89050Gent,BelgiumEditor:Mikio BraunAbstractWhen analyzing data from computationally expensive simulation codes,surrogate model-ing methods arefirmly established as facilitators for design space exploration,sensitivity analysis,visualization and optimization.Kriging is a popular surrogate modeling tech-nique used for the Design and Analysis of Computer Experiments(DACE).Hence,the past decade Kriging has been the subject of extensive research and many extensions have been proposed,e.g.,co-Kriging,stochastic Kriging,blind Kriging,etc.However,few Krig-ing implementations are publicly available and tailored towards scientists and engineers.Furthermore,no Kriging toolbox exists that unifies several Krigingflavors.This paper addresses this need by presenting an efficient object-oriented Kriging implementation and several Kriging extensions,providing aflexible and easily extendable framework to test and implement new Krigingflavors while reusing as much code as possible.Keywords:Kriging,Gaussian process,co-Kriging,blind Kriging,surrogate modeling, metamodeling,DACE1.IntroductionThis paper is concerned with efficiently solving complex,computational expensive problems using surrogate modeling techniques(Gorissen et al.,2010).Surrogate models,also known as metamodels,are cheap approximation models for computational expensive(black-box) simulations.Surrogate modeling techniques are well-suited to handle,for example,expen-sivefinite element(FE)simulations and computationalfluid dynamic(CFD)simulations.Kriging is a popular surrogate model type to approximate deterministic noise-free data. First conceived by Danie Krige in geostatistics and later introduced for the Design and Analysis of Computer Experiments(DACE)by Sacks et al.(1989),these Gaussian pro-cess(Rasmussen and Williams,2006)based surrogate models are compact and cheap to evaluate,and have proven to be very useful for tasks such as optimization,design space exploration,visualization,prototyping,and sensitivity analysis(Viana et al.,2014).Note ∗.Ivo Couckuyt is a post-doctoral research fellow of FWO-Vlaanderen.Couckuyt,Dhaene and Demeesterthat Kriging surrogate models are primarily known as Gaussian processes in the machine learning community.Except for the utilized terminology there is no difference between the terms and associated methodologies.While Kriging is a popular surrogate model type,not many publicly available,easy-to-use Kriging implementations exist.Many Kriging implementations are outdated and often limited to one specific type of Kriging.Perhaps the most well-known Kriging toolbox is the DACE toolbox1of Lophaven et al.(2002),but,unfortunately,the toolbox has not been updated for some time and only the standard Kriging model is provided.Other freely available Kriging codes include:stochastic Kriging(Staum,2009),2DiceKriging,3 Gaussian processes for Machine Learning(Rasmussen and Nickisch,2010)(GPML),4demo code provided with Forrester et al.(2008),5and the Matlab Krigeage toolbox.6 This paper addresses this need by presenting an object-oriented Kriging implementation and several Kriging extensions,providing aflexible and easily extendable framework to test and implement new Krigingflavors while reusing as much code as possible.2.ooDACE ToolboxThe ooDACE toolbox is an object-oriented Matlab toolbox implementing a variety of Krig-ingflavors and extensions.The most important features and Krigingflavors include:•Simple Kriging,ordinary Kriging,universal Kriging,stochastic Kriging(regression Kriging),blind-and co-Kriging.•Derivatives of the prediction and prediction variance.•Flexible hyperparameter optimization.•Useful utilities include:cross-validation,integrated mean squared error,empirical variogram plot,debug plot of the likelihood surface,robustness-criterion value,etc.•Proper object-oriented design(compatible interface with the DACE toolbox1is avail-able).Documentation of the ooDACE toolbox is provided in the form of a getting started guide (for users),a wiki7and doxygen documentation8(for developers and more advanced users). In addition,the code is well-documented,providing references to research papers where appropriate.A quick-start demo script is provided withfive surrogate modeling use cases, as well as script to run a suite of regression tests.A simplified UML class diagram,showing only the most important public operations, of the toolbox is shown in Figure1.The toolbox is designed with efficiency andflexibil-ity in mind.The process of constructing(and predicting)a Kriging model is decomposed in several smaller,logical steps,e.g.,constructing the correlation matrix,constructing the1.The DACE toolbox can be downloaded at http://www2.imm.dtu.dk/~hbn/dace/.2.The stochastic Kriging toolbox can be downloaded at /.3.The DiceKriging toolbox can be downloaded at /web/packages/DiceKriging/index.html.4.The GPML toolbox can be downloaded at /software/view/263/.5.Demo code of Kriging can be downloaded at //legacy/wileychi/forrester/.6.The Krigeage toolbox can be downloaded at /software/kriging/.7.The wiki documentation of the ooDACE toolbox is found at http://sumowiki.intec.ugent.be/index.php/ooDACE:ooDACE_toolbox.8.The doxygen documentation of the ooDACE toolbox is found at http://sumo.intec.ugent.be/buildbot/ooDACE/doc/.Figure1:Class diagram of the ooDACE toolbox.regression matrix,updating the model,optimizing the parameters,etc.These steps are linked together by higher-level steps,e.g.,fitting the Kriging model and making predic-tions.The basic steps needed for Kriging are implemented as(protected)operations in the BasicGaussianProcess superclass.Implementing a new Kriging type,or extending an existing one,is now done by subclassing the Kriging class of your choice and inheriting the(protected)methods that need to be reimplemented.Similarly,to implement a new hyperparameter optimization strategy it suffices to create a new class inherited from the Optimizer class.To assess the performance of the ooDACE toolbox a comparison between the ooDACE toolbox and the DACE toolbox1is performed using the2D Branin function.To that end,20data sets of increasing size are constructed,each drawn from an uniform random distribution.The number of observations ranges from10to200samples with steps of10 samples.For each data set,a DACE toolbox1model,a ooDACE ordinary Kriging and a ooDACE blind Kriging model have been constructed and the accuracy is measured on a dense test set using the Average Euclidean Error(AEE).Moreover,each test is repeated 1000times to remove any random factor,hence the average accuracy of all repetitions is used.Results are shown in Figure2a.Clearly,the ordinary Kriging model of the ooDACE toolbox consistently outperforms the DACE toolbox for any given sample size,mostly due to a better hyperparameter optimization,while the blind Kriging model is able improve the accuracy even more.3.ApplicationsThe ooDACE Toolbox has already been applied successfully to a wide range of problems, e.g.,optimization of a textile antenna(Couckuyt et al.,2010),identification of the elasticity of the middle-ear drum(Aernouts et al.,2010),etc.In sum,the ooDACE toolbox aims to provide a modern,up to date Kriging framework catered to scientists and age instructions,design documentation,and stable releases can be found at http://sumo.intec.ugent.be/?q=ooDACE.ReferencesJ.Aernouts,I.Couckuyt,K.Crombecq,and J.J.J.Dirckx.Elastic characterization of membranes with a complex shape using point indentation measurements and inverseCouckuyt,Dhaene and Demeester(a)(b)Figure2:(a)Evolution of the average AEE versus the number of samples(Branin function).(b)Landscape plot of the Branin function.modelling.International Journal of Engineering Science,48:599–611,2010.I.Couckuyt,F.Declercq,T.Dhaene,and H.Rogier.Surrogate-based infill optimization applied to electromagnetic problems.Journal of RF and Microwave Computer-Aided Engineering:Advances in Design Optimization of Microwave/RF Circuits and Systems, 20(5):492–501,2010.A.Forrester,A.Sobester,and A.Keane.Engineering Design Via Surrogate Modelling:A Practical Guide.Wiley,Chichester,2008.D.Gorissen,K.Crombecq,I.Couckuyt,P.Demeester,and T.Dhaene.A surrogate modeling and adaptive sampling toolbox for computer based design.Journal of Machine Learning Research,11:2051–2055,2010.URL http://sumo.intec.ugent.be/.S.N.Lophaven,H.B.Nielsen,and J.Søndergaard.Aspects of the Matlab toolbox DACE. Technical report,Informatics and Mathematical Modelling,Technical University of Den-mark,DTU,Richard Petersens Plads,Building321,DK-2800Kgs.Lyngby,2002.C.E.Rasmussen and H.Nickisch.Gaussian processes for machine learning(GPML)toolbox. Journal of Machine Learning Research,11:3011–3015,2010.C.E.Rasmussen and C.K.I.Williams.Gaussian Processes for Machine Learning.MIT Press,2006.J.Sacks,W.J.Welch,T.J.Mitchell,and H.P.Wynn.Design and analysis of computer experiments.Statistical Science,4(4):409–435,1989.J.Staum.Better simulation metamodeling:The why,what,and how of stochastic Kriging. In Proceedings of the Winter Simulation Conference,2009.F.A.C.Viana,T.W.Simpson,V.Balabanov,and V.Toropov.Metamodeling in multi-disciplinary design optimization:How far have we really come?AIAA Journal,52(4): 670–690,2014.。
AVVIO RAPIDO Router mobili 5G 安装指南说明书
Per caricare la batteria, collegare il cavo USB al router mobile, quindi collegarlo a una presa a muro utilizzando l'adattatore di alimentazione CA o una porta USB del computer.Assicurarsi che l'orientamento della scheda nano SIM coincida con l'orientamento indicato sull'etichetta del dispositivo e inserirla delicatamente, quindi posizionare la batteria e il coperchio posteriore.NOTA: utilizzare solo le dita per inserire o rimuovere la scheda nano SIM. L'utilizzo di altri oggetti potrebbe danneggiare il dispositivo.1. COM'È FATTO IL DISPOSITIVO2. INSTALLAZIONE DELLA SIM E DELLA BATTERIAIl router mobile viene fornito con i seguenti componenti:• Router mobile Nighthawk® M6 o M6 Pro 5G*• Coperchio della batteria • Batteria• Cavo USB Tipo C• Alimentatore (varia in base all’area geografica)• Adattatori con presa Tipo C (per la maggior parte dei Paesi europei)•Adattatori con presa Tipo G (per il Regno Unito)*Illustrazioni del modello Nighthawk M6 per scopi illustrativi.antenna esterna (TS-9)antenna esterna (TS-9)USB Tipo CEthernetCONFORMITÀ NORMATIVA E NOTE LEGALIPer informazioni sulla conformità alle normative, compresala Dichiarazione di conformità UE, visitare il sito Web https:///it/about/regulatory/.Prima di collegare l'alimentazione, consultare il documento relativo alla conformità normativa.Può essere applicato solo ai dispositivi da 6 GHz: utilizzare il dispositivo solo in un ambiente al chiuso. L'utilizzo di dispositivi a 6 GHz è vietato su piattaforme petrolifere, automobili, treni, barche e aerei, tuttavia il suo utilizzo è consentito su aerei di grandi dimensioni quando volano sopra i 3000 metri di altezza. L'utilizzo di trasmettitori nella banda 5.925‑7.125 GHz è vietato per il controllo o le comunicazioni con sistemi aerei senza equipaggio.SUPPORTO E COMMUNITYDalla pagina del portale di amministrazione Web, fare clic sull'icona con i tre puntini nell'angolo in alto a destra per accedere ai file della guida e del supporto.Per ulteriori informazioni, visitare il sito netgear.it/support per accedere al manuale dell'utente completo e per scaricare gli aggiornamenti del firmware.È possibile trovare utili consigli anche nella Community NETGEAR, alla pagina /it.GESTIONE DELLE IMPOSTAZIONI TRAMITE L'APP NETGEAR MOBILEUtilizzare l'app NETGEAR Mobile per modificare il nome della rete Wi-Fi e la password. È possibile utilizzarla anche per riprodurre e condividere contenutimultimediali e accedere alle funzioni avanzate del router mobile.1. Accertarsi che il dispositivo mobile sia connesso a Internet.2. Eseguire la scansione del codice QR per scaricare l'appNETGEAR Mobile.Connessione con il nome e la password della rete Wi-Fi 1. Aprire il programma di gestione della rete Wi‑Fi deldispositivo.2. Individuare il nome della rete Wi‑Fi del router mobile(NTGR_XXXX) e stabilire una connessione.3. Only Connessione tramite EthernetPer prolungare la durata della batteria, l'opzione Ethernet è disattivata per impostazione predefinita. Per attivarla, toccare Power Manager (Risparmio energia) e passare a Performance Mode (Modalità performance).4. CONNESSIONE A INTERNETÈ possibile connettersi a Internet utilizzando il codice QR del router mobile da uno smartphone oppure selezionando manualmente il nome della rete Wi‑Fi del router e immettendo la password.Connessione tramite codice QR da uno smartphone 1. Toccare l'icona del codice QR sulla schermata inizialedello schermo LCD del router mobile.NOTA: quando è inattivo, lo schermo touch si oscura per risparmiare energia. Premere brevemente e rilasciare il pulsante di alimentazione per riattivare lo schermo.3. CONFIGURAZIONE DEL ROUTER MOBILETenere premuto il pulsante di accensione per due secondi, quindi seguire le istruzioni visualizzate sullo schermo per impostare un nome per la rete Wi‑Fi e una password univoci.La personalizzazione delle impostazioni Wi‑Fi consente di proteggere la rete Wi‑Fi del router mobile.Impostazioni APNIl router mobile legge i dati dalla scheda SIM e determina automaticamente le impostazioni APN (Access Point Name) corrette con i piani dati della maggior parte degli operatori. Tuttavia, se si utilizza un router mobile sbloccato con un operatore o un piano meno comune, potrebbe essere necessario immettere manualmente le impostazioni APN.Se viene visualizzata la schermata APN Setup Required (Configurazione APN richiesta), i dati APN dell’operatore non sono presenti nel nostro database ed è necessario inserirli manualmente. Immettere i valori fornitidall’operatore nei campi corrispondenti, quindi toccare Save (Salva) per completare la configurazione.NOTA: l’operatore determina le proprie informazioni APN e deve fornire le informazioni per il proprio piano dati. Si consiglia di contattare il proprio operatore per le impostazioni APN corrette e di utilizzare solo l’APN suggerito per il piano specifico.Schermata inizialeAl termine della configurazione, il router visualizza la schermata iniziale:Wi‑FiPotenza Carica Rete Codice QR connessione rapida Wi‑FiNome e Wi‑FiIcona del codice QR。
布莱兹帕斯卡与数学概率的应用
布莱兹帕斯卡与数学概率的应用布莱兹·帕斯卡（Blaise Pascal）是一位杰出的法国数学家、物理学家和哲学家，他不仅在数学领域作出了重要贡献，同时也是概率论的奠基人之一。
帕斯卡在17世纪的工作，不仅推动了概率论的发展，也为后来的统计学奠定了基础。
本文将探讨帕斯卡在概率论方面的突出贡献，以及这些理论在实际生活和现代科学中的应用。
帕斯卡的生平与背景布莱兹·帕斯卡于1623年出生在法国克兰，父亲是当地的一位地方法官，母亲早逝。
尽管年幼失去了母亲，他在父亲的指导下表现出超常的智力和才华。
年仅12岁时，帕斯卡便开始学习几何，并很快展现出对数学理性的独特理解。
帕斯卡的一生充满了对知识的追求。
他在科学、技术和哲学等多个领域都做出了卓越贡献，并积累了丰富的研究成果。
他在仅享年39岁时去世，但他的思想与学术成就仍然深远地影响着后人，尤其是在数学和概率论上。
概率论的起源概率论并不是某一个单一数学家的发明，而是在多个历史阶段逐渐形成的一门科学。
然而，帕斯卡与另一位数学家费马（Pierre de Fermat）的通信，被普遍认为是现代概率论的开端。
在1654年，二人就“赌博问题”进行了深入探讨，这些问题主要涉及如何计算在不完整信息或不确定条件下获胜的机会。
例如，在扔骰子或抽牌等游戏中，参与者需要动态评估各自获胜的可能性。
基于这些问题，帕斯卡与费马通过数理逻辑形式化了这些概念。
这不仅让游戏更具趣味性，也开创了评估随机事件结果概率的方法。
帕斯卡的主要贡献1. 概率的定义帕斯卡和费马对概率的初步讨论引入了随机事件中可能结果计算的方法。
他们强调，在面对不确定性时，概率可以被作为一种理性决策工具。
使用简单但有效的方法，他们将概率定义为成功事件数与所有可能事件数之比。
这一定义成为后续研究中的重要基石。
2. 投机会话题“赌博问题”成为了帕斯卡概率论发展的一个重要分水岭，其中最著名的是“骰子问题”。
当两个玩家在游戏中由于时间原因中止比赛时，如何合理分配未完成游戏中的胜利机会？帕斯卡通过对每个球员获胜情况及其相应概率进行分析，设计出数学模型来解决问题。
Backward stochastic differential equations in finance
where f is the generator and ξ is the terminal condition. Actually, this type of equation appears in numerous problems in finance (as pointed out in Quenez’s doctorate 1993). First, the theory of contingent claim valuation in a complete market studied by Black and Scholes (1973), Merton (1973, 1991), Harrison and Kreps (1979), Harrison and Pliska (1981), Duffie (1988), and Karatzas (1989), among others, can be expressed in terms of BSDEs. Indeed, the problem is to determine the price of a contingent claim ξ ≥ 0 of maturity T , which is a contract that pays an amount ξ at time T . In a complete market it is possible to construct a portfolio which attains as final wealth the amount ξ . Thus, the dynamics of the value of the replicating portfolio Y are given by a BSDE with linear generator f , with Z corresponding to the hedging portfolio. Then the price at time t is associated naturally with the value at time t of the hedging portfolio. However, there exists an infinite number of replicating portfolios and consequently the
科勒卫浴保修卡说明书
INODOROS Y MUEBLES DE BAÑO El presente certificado cubre cualquier defecto de fabricación del producto que pueda afectar su desempeño y en cumplimiento de lo establecido en el manual de instalación, uso y mantenimiento.CENTRO CERAMICO LAS FLORES SAC, certifica la garantía de los sanitarios de la marca Origin, Golden Bath, Bravat, Trébol Platinum y Noken según las siguientes condiciones: presentación del documento de venta, factura o boleta (original o virtual).2.Si presenta defectos de fabricación del producto y no por fallas que se presenten como consecuencia de lainadecuada instalación, reparación, uso o un mal mantenimiento. Garantía sólo cubre el valor del componente a cambiar, si fuese necesario se hará cambio total del producto, no cubre mano de obra traslado, ni otros gastos por remover o instalar productos.4.En caso se reportara alguna deficiencia del sanitario o mueble, comunicarse con el área de Servicio Técnico de CENTRO CERAMICO LAS FLORES SAC quien verificará a fin de establecer la idoneidad y calidad del producto. Garantía no contempla rajaduras ni roturas originadas por golpe que se den en el transporte y/o almacenaje inadecuado.Es muy importante que revise su mercadería al momento de recibirla dando su conformidad de la misma. Garantía fuera de Lima Metropolitana aplica en la ciudades donde se encuentran las tiendas Cassinelli; en otras ciudades la visita técnica a solicitud del cliente se realizará previa cotización y la validación del cliente a asumir el costo que demande la visita hacia la ciudad donde se encuentre instalado el producto. En caso el producto presentara falla de fabricación se aplicará lagarantía según el punto número 3, brindando la atención en tienda.7.No colocar peso excesivo y/o subirse sobre el mueble y tampoco sobre los inodoros.8.No exponer el producto directamente al sol.9.Evite contacto directo del agua o cualquier otro liquido sobre el mueble de baño para evitar la filtración en el interior de la estructura y se produzca hinchamiento.10.Evitar golpear bruscamente al cerrar las puertas y cajones.11.Evitar arrojar papeles dentro del inodoro.Tiempos de Cobertura:El presente certificado de garantía para one pieces, inodoros, bowls,lavatorios y muebles de baño tendrán vigencia a partir de la fecha de compra del producto indicado en su documento de compra.INFORMACION A CONSIDERARLos sanitarios presentarán curaciones ( parches ) en algunas zonas no visibles de la loza (inodoros o lavatorios).En el proceso de fabricación a altas temperaturas se hallan orificios que son necesarios para la ventilación y pequeñas fisuras que se dan en el proceso de enfriamiento tanto en la base interna y tanque del inodoro e igual forma por debajo del lavatorio.Tener en consideración que la loza sanitaria puede presentar defectos de esmalte permisible según norma técnica de fabricación.no debiendo ser visible a más de 1 metro de distancia de la zona frontal y lateral del producto.TREBOL PLATINUM ORIGIN De por Vida (defecto Fábrica)Accesorio interno (3 años)Complementos (1 año)GOLDEN BATH De por Vida (defecto Fábrica)Accesorio interno (1 año) Complementos(6 meses)3 años1 año Accesorio interno (3 años)Complementos (1 año)De por Vida (defecto Fábrica)LAVATARIOS Y MUEBLESMueble Golden Bath 3 años 6 mesesMueble Trébol Platinum 3 años1 año Mueble Origin CERTIFICADO DE GARANTIASANITARIOS Y MUEBLES DE BAÑOBRAVAT MARCA TABLEROGABINETE LOZA SANITARIA MARCA 2 años10 años NOKEN TIEMPOS DE GARANTIA DE SANITARIOS1 año5 años ACCESORIO Y COMPLEMENTOInspiramos el cambio en tuCERTIFICADO DE GARANTIA INODOROS Y MUEBLES DE BAÑOINSTALACION 1.Contratar personal calificado para la instalación de los productos adquiridos.2.Revisar los componentes del producto antes de la instalación (hacer lectura del manual propio de instalación o video brindado).3.Antes de instalar el inodoro se debe verificar si la medida del eje del punto de desagüe de piso o pared va acorde con la del inodoro,asimismo considerar la medida del punto de agua.4.El tubo de desagüe y la toma de agua, tanto interior como exterior, deben encontrarse libres de impurezas y residuos de cemento que puedan atorar el drenaje de agua.5.Para una instalación de inodoro al piso se debe marcar la posición de los pernos de anclaje en la superficie de forma paralela a la pared terminada, posteriormente perforar e instalar los pernos.6.Colocar el anillo de cera alrededor del desagüe del inodoro luego asentar el inodoro al piso y se asegura en la posición definitiva colocando los pernos de anclaje y ajustando a la loza. El anillo debe formar un sello entre el piso y el inodoro para evitar la salida de malos olores y filtraciones de agua. Hacer pruebas de funcionamiento antes de sellar. fijación del borde de la taza al piso o lavatorios de loza, debe realizarse con alguna silicona neutra para uso sanitario.Este producto permite desmontar la loza sin quebrarla (no usar productos que imposibiliten el desmontaje)MANTENIMIENTOSANITARIOS*Para la limpieza de la loza sanitaria no debe usarse materiales abrasivos como escobilla de alambre o similares que pueden deteriorar el acabado de la loza sanitaria.*Combinar los químicos de limpieza con agua antes de su aplicación para evitar dañar el acabado cromado de las bisagras decorativas,pulsadores o el mismo asiento.MUEBLES DE BAÑO*Los lavatorios de loza deben ser limpiados con agua y jabón líquido utilizando una esponja suave.*Posteriormente secar con un paño para evitar manchas.*Evitar acumulación de agua sobre el mueble, podría generar hinchamiento en su estructura..*Limpiar el mueble únicamente con paño ligeramente húmedo.*No aplicar detergentes y evitar la limpieza con esponjas abrasivas ya sea en el lavatorio y el propio mueble de baño.*Es importante la limpieza diaria de la loza para evitar manchas en la superficie (se recomienda utilizar bicarbonato de sodio y vinagre blanco para eliminar la mancha en la loza).Los datos personales que usted proporciona serán utilizados y/o tratados por Centro Cerámico Las Flores SAC estricta y únicamente a efectos de brindarle atención personalizada para la gestión de una posible solución del inconveniente reportado, así como para la acreditación de la atención del mismo. Centro Cerámico Las Flores SAC podrá compartir y/o usar y/o almacenar y/o transferir su información a terceras personas vinculadas o no a Centro Cerámico Las Flores SAC. sean estos socios comerciales o no de Centro Cerámico Las Flores SAC, con el objeto de realizar las actividades relacionadas a la atención post ventas y/o servicio técnico solicitado. Usted podrá ejercer en cualquier momento su derecho de información, acceso, rectificación, cancelación y oposición de sus datos de acuerdo a lo dispuesto por la Ley de Protección de Datos Personales, vigente y su Reglamento. Para más información en la web ó por correo a ********************************.Nombre de Titular de Compra ………………………………………………………………………………………………………………………………………………………………………..Numero de Boleta o Factura ………………………………………....……………………………………………………….Fecha de Compra …………………………………….DNI ………………………………………………………………………………………………………………………………..……….Firma ………………………………………………………….Nombre de Contacto ………………………………………………………………………………………………………..…….Firma de Contacto ………………………………………INSTALACION Y MANTENIMIENTO DE INODOROS Y MUEBLES DE BAÑOServicio Técnico CCLFTeléfono 0800-1-2150 / 241-5753 / 243-2242/ 241-5746Contáctenosvíaweb:******************************************************Inspiramos el cambio en tu hogar INSTALACION Y MATENIMIENTODE INODOROS Y MUEBLES DE BAÑO。
Some Recent Aspects of Differential Game Theory
Dyn Games Appl(2011)1:74–114DOI10.1007/s13235-010-0005-0Some Recent Aspects of Differential Game TheoryR.Buckdahn·P.Cardaliaguet·M.QuincampoixPublished online:5October2010©Springer-Verlag2010Abstract This survey paper presents some new advances in theoretical aspects of dif-ferential game theory.We particular focus on three topics:differential games with state constraints;backward stochastic differential equations approach to stochastic differential games;differential games with incomplete information.We also address some recent devel-opment in nonzero-sum differential games(analysis of systems of Hamilton–Jacobi equa-tions by conservation laws methods;differential games with a large number of players,i.e., mean-field games)and long-time average of zero-sum differential games.Keywords Differential game·Viscosity solution·System of Hamilton–Jacobi equations·Mean-field games·State-constraints·Backward stochastic differential equations·Incomplete information1IntroductionThis survey paper presents some recent results in differential game theory.In order to keep the presentation at a reasonable size,we have chosen to describe in full details three topics with which we are particularly familiar,and to give a brief summary of some other research directions.Although this choice does not claim to represent all the recent literature on the R.Buckdahn·M.QuincampoixUniversitéde Brest,Laboratoire de Mathématiques,UMR6205,6Av.Le Gorgeu,BP809,29285Brest, FranceR.Buckdahne-mail:Rainer.Buckdahn@univ-brest.frM.Quincampoixe-mail:Marc.Quincampoix@univ-brest.frP.Cardaliaguet( )Ceremade,UniversitéParis-Dauphine,Place du Maréchal de Lattre de Tassigny,75775Paris Cedex16, Francee-mail:cardaliaguet@ceremade.dauphine.frmore theoretic aspects of differential game theory,we are pretty much confident that it cov-ers a large part of what has recently been written on the subject.It is clear however that the respective part dedicated to each topic is just proportional to our own interest in it,and not to its importance in the literature.The three main topics we have chosen to present in detail are:–Differential games with state constraints,–Backward stochastic differential equation approach to differential games,–Differential games with incomplete information.Before this,we also present more briefly two domains which have been the object of very active research in recent years:–nonzero-sum differential games,–long-time average of differential games.Thefirst section of this survey is dedicated to nonzero-sum differential games.Although zero-sum differential games have attracted a lot of attention in the80–90’s(in particular, thanks to the introduction of viscosity solutions for Hamilton–Jacobi equations),the ad-vances on nonzero-sum differential games have been scarcer,and mainly restricted to linear-quadratic games or stochastic differential games with a nondegenerate diffusion.The main reason for this is that there was very little understanding of the system of Hamilton–Jacobi equations naturally attached to these games.In the recent years the analysis of this sys-tem has been the object of several papers by Bressan and his co-authors.At the same time, nonzero-sum differential games with a very large number of players have been investigated in the terminology of mean-field games by Lasry and Lions.In the second section we briefly sum up some advances in the analysis of the large time behavior of zero-sum differential games.Such problems have been the aim of intense re-search activities in the framework of repeated game theory;it has however only been re-cently investigated for differential games.In the third part of this survey(thefirst one to be the object of a longer development) we investigate the problem of state constraints for differential games,and in particular,for pursuit-evasion games.Even if such class of games has been studied since Isaacs’pioneer-ing work[80],the existence of a value was not known up to recently for these games in a rather general framework.This is mostly due to the lack of regularity of the Hamiltonian and of the value function,which prevents the usual viscosity solution approach to work(Evans and Souganidis[63]):Indeed some controllability conditions on the phase space have to be added in order to prove the existence of the value(Bardi,Koike and Soravia[18]).Following Cardaliaguet,Quincampoix and Saint Pierre[50]and Bettiol,Cardaliaguet and Quincam-poix[26]we explain that,even without controllability conditions,the game has a value and that this value can be characterized as the smallest supersolution of some Hamilton–Jacobi equation with discontinuous Hamiltonian.Next we turn to zero-sum stochastic differential games.Since the pioneering work by Fleming and Souginidis[65]it has been known that such games have a value,at least in a framework of games of the type“nonanticipating strategies against controls”.Unfortunately this notion of strategies is not completely satisfactory,since it presupposes that the players have a full knowledge of their opponent’s control in all states of the world:It would be more natural to assume that the players use strategies which give an answer to the control effectively played by their opponent.On the other hand it seems also natural to consider nonlinear cost functionals and to allow the controls of the players to depend on events of the past which happened before the beginning of the game.The last two points have beeninvestigated in a series of papers by Buckdahn and Li[35,36,39],and an approach more direct than that in[65]has been developed.Thefirst point,together with the two others,will be the object of the fourth part of the survey.In the last part we study differential games with incomplete information.In such games, one of the parameters of the game is chosen at random according to some probability mea-sure and the result is told to one of the players and not to the other.Then the game is played as usual,players observing each other’s control.The main difference with the usual case is that at least one of the players does not know which payoff he is actually optimizing.All the difficulty of this game is to understand what kind of information the informed player has interest in to disclose in order to optimize his payoff,taking thus the risk that his opponent learns his missing information.Such games are the natural extension to differential games of the Aumann–Maschler theory for repeated games[11].Their analysis has been developed in a series of papers by Cardaliaguet[41,43–45]and Cardaliaguet and Rainer[51,52].Throughout these notes we assume the reader to be familiar with the basic results of dif-ferential game theory.Many references can be quoted on this subject:A general introduction for the formal relation between differential games and Hamilton–Jacobi equations(or sys-tem)can be found in the monograph Baçar and Olsder[13].We also refer the reader to the classical monographs by Isaacs[80],Friedman[67]and Krasovskii and Subbotin[83]for early presentations of differential game theory.The recent literature on differential games strongly relies on the notion of viscosity solution:Classical monographs on this subject are Bardi and Capuzzo Dolcetta[17],Barles[19],Fleming and Soner[64],Lions[93]and the survey paper by Crandall,Ishii and Lions[56].In particular[17]contains a good introduc-tion to the viscosity solution aspects of deterministic zero-sum differential games:the proof of the existence and the characterization of a value for a large class of differential games can be found there.Section6is mostly based on the notion of backward stochastic differential equation(BSDE):We refer to El Karoui and Mazliak[60],Ma and Yong[96]and Yong and Zhou[116]for a general presentation.The reader is in particular referred to the work by S.Peng on BSDE methods in stochastic control[101].Let usfinally note that,even if this survey tries to cover a large part of the recent literature on the more theoretical aspects of differential games,we have been obliged to omit some topics:linear-quadratic differential games are not covered by this survey despite their usefulness in applications;however,these games have been already the object of several survey ck of place also prevented us from describing advances in the domain of Dynkin games.2Nonzero-sum Differential GamesIn the recent years,the more striking advances in the analysis of nonzero-sum differential games have been directed in two directions:analysis by P.D.E.methods of Nash feedback equilibria for deterministic differential games;differential games with a very large number of small players(mean-field games).These topics appear as the natural extensions of older results:existence of Nash equilibria in memory strategies and of Nash equilibria in feedback strategies for stochastic differential games,which have also been revisited.2.1Nash Equilibria in Memory StrategiesSince the work of Kononenko[82](see also Kleimenov[81],Tolwinski,Haurie and Leit-mann[114],Gaitsgory and Nitzan[68],Coulomb and Gaitsgory[55]),it has been knownthat deterministic nonzero-sum differential games admit Nash equilibrium payoffs in mem-ory strategies:This result is actually the counterpart of the so-called Folk Theorem in re-peated game theory[100].Recall that a memory(or a nonanticipating)strategy for a player is a strategy where this player takes into account the past controls played by the other play-ers.In contrast a feedback strategy is a strategy which only takes into account the present position of the system.Following[82]Nash equilibrium payoffs in memory strategies are characterized as follows:A payoff is a Nash equilibrium payoff if and only if it is reach-able(i.e.,the players can obtain it by playing some control)and individually rational(the expected payoff for a player lies above its min-max level at any point of the resulting trajec-tory).This result has been recently generalized to stochastic differential games by Buckdahn, Cardaliaguet and Rainer[38](see also Rainer[105])and to games in which players can play random strategies by Souquière[111].2.2Nash Equilibria in Feedback FormAlthough the existence and characterization result of Nash equilibrium payoffs in mem-ory strategies is quite general,it has several major drawbacks.Firstly,there are,in general, infinitely many such Nash equilibria,but there exists—at least up to now—no completely satisfactory way to select one.Secondly,such equilibria are usually based on threatening strategies which are often non credible.Thirdly,the corresponding strategies are,in general, not“time-consistent”and in particular cannot be computed by any kind of“backward in-duction”.For this reason it is desirable tofind more robust notions of Nash equilibria.The best concept at hand is the notion of subgame perfect Nash equilibria.Since the works of Case[54]and Friedman[67],it is known that subgame perfect Nash equilibria are(at least heuristically)given by feedback strategies and that their corresponding payoffs should be the solution of a system of Hamilton–Jacobi equations.Up to now these ideas have been successfully applied to linear-quadratic differential games(Case[54],Starr and Ho[113], ...)and to stochastic differential games with non degenerate viscosity term:In thefirst case,one seeks solutions which are quadratic with respect to the state variable;this leads to the resolution of Riccati equations.In the latter case,the regularizing effect of the non-degenerate diffusion allows us to usefixed point arguments to get either Nash equilibrium payoffs or Nash equilibrium feedbacks.Several approaches have been developed:Borkar and Ghosh[27]consider infinite horizon problems and use the smoothness of the invari-ant measure associated to the S.D.E;Bensoussan and Frehse[21,22]and Mannucci[97] build“regular”Nash equilibrium payoffs satisfying a system of Hamilton–Jacobi equations thanks to elliptic or parabolic P.D.E techniques;Nash equilibrium feedbacks can also be built by backward stochastic differential equations methods like in Hamadène,Lepeltier and Peng[75],Hamadène[74],Lepeltier,Wu and Yu[92].2.3Ill-posedness of the System of HJ EquationsIn a series of articles,Bressan and his co-authors(Bressan and Chen[33,34],Bressan and Priuli[32],Bressan[30,31])have analyzed with the help of P.D.E methods the system of Hamilton–Jacobi equations arising in the construction of feedback Nash equilibria for deter-ministic nonzero-sum games.In state-space dimension1and for thefinite horizon problem, this system takes the form∂V i+H i(x,D V1,...,D V n)=0in R×(0,T),i=1,...,n,coupled with a terminal condition at time T(here n is the number of players and H i is the Hamiltonian of player i,V i(t,x)is the payoff obtained by player i for the initial condition (t,x)).Setting p i=(V i)x and deriving the above system with respect to x one obtains the system of conservation laws:∂t p i+H i(x,p1,...,p n)x=0in R×(0,T).This system turns out to be,in general,ill-posed.Typically,in the case of two players(n= 2),the system is ill-posed if the terminal payoff of the players have an opposite monotonicity. If,on the contrary,these payoffs have the same monotony and are close to some linear payoff (which is a kind of cooperative case),then the above system has a unique solution,and one can build Nash equilibria in feedback form from the solution of the P.D.E[33].Still in space dimension1,the case of infinite horizon seems more promising:The sys-tem of P.D.E then reduces to an ordinary differential equation.The existence of suitable solutions for this equation then leads to Nash equilibria.Such a construction is carried out in Bressan and Priuli[32],Bressan[30,31]through several classes of examples and by various methods.In a similar spirit,the papers Cardaliaguet and Plaskacz[47],Cardaliaguet[42]study a very simple class of nonzero-sum differential games in dimension1and with a terminal payoff:In this case it is possible to select a unique Nash equilibrium payoff in feedback form by just imposing that it is Pareto whenever there is a unique Pareto one.However,this equilibrium payoff turns out to be highly unstable with respect to the terminal data.Some other examples of nonlinear-quadratic differential games are also analyzed in Olsder[99] and in Ramasubramanian[106].2.4Mean-field GamesSince the system of P.D.Es arising in nonzero-sum differential games is,in general,ill-posed,it is natural to investigate situations where the problem simplifies.It turns out that this is the case for differential games with a very large number of identical players.This problem has been recently developed in a series of papers by Lasry and Lions[87–90,94] under the terminology of mean-field games(see also Huang,Caines and Malhame[76–79] for a related approach).The main achievement of Lasry and Lions is the identification of the limit when the number of players tends to infinity.The typical resulting model takes the form⎧⎪⎨⎪⎩(i)−∂t u−Δu+H(x,m,Du)=0in R d×(0,T),(ii)∂t m−Δm−divD p H(x,m,Du)m=0in R d×(0,T),(iii)m(0)=m0,u(x,T)=Gx,m(T).(1)In the above system,thefirst equation has to be understood backward in time while the second one is forward in time.Thefirst equation(a Hamilton–Jacobi one)is associated with an optimal control problem and its solution can be regarded as the value function for a typical small player(in particular the Hamiltonian H=H(x,m,p)is convex with respect to the last variable).As for the second equation,it describes the evolution of the density m(t)of the population.More precisely,let usfirst consider the behavior of a typical player.He controls through his control(αs)the stochastic differential equationdX t=αt dt+√2B t(where(B t)is a standard Brownian motion)and he aims at minimizing the quantityET12LX s,m(s),αsds+GX T,m(T),where L is the Fenchel conjugate of H with respect to the p variable.Note that in this cost the evolving measure m(s)enters as a parameter.The value function of our average player is then given by(1-(i)).His optimal control is—at least heuristically—given in feedback form byα∗(x,t)=−D p H(x,m,Du).Now,if all agents argue in this way,their repartition will move with a velocity which is due,on the one hand,to the diffusion,and,one the other hand,to the drift term−D p H(x,m,Du).This leads to the Kolmogorov equation(1-(ii)).The mean-field game theory developed so far has been focused on two main issues:firstly,investigate equations of the form(1)and give an interpretation(in economics,for instance)of such systems.Secondly,analyze differential games with afinite but large num-ber of players and interpret(1)as their limiting behavior as the number of players goes to infinity.Up to now thefirst issue is well understood and well documented.The original works by Lasry and Lions give a certain number of conditions under which(1)has a solution,discuss its uniqueness and its stability.Several papers also study the numerical approximation of this solution:see Achdou and Capuzzo Dolcetta[1],Achdou,Camilli and Capuzzo Dolcetta[2], Gomes,Mohr and Souza[71],Lachapelle,Salomon and Turinici[85].The mean-field games theory has been used in the analysis of wireless communication systems in Huang,Caines and Malhamé[76],or Yin,Mehta,Meyn and Shanbhag[115].It seems also particularly adapted to modeling problems in economics:see Guéant[72,73],Lachapelle[84],Lasry, Lions,Guéant[91],and the references therein.As for the second part of the program,the limiting behavior of differential games when the number of players tend to infinity has been understood for ergodic differential games[88].The general case remains mostly open.3Long-time Average of Differential GamesAnother way to reduce the complexity of differential games is to look at their long-time be-havior.Among the numerous applications of this topic let us quote homogenization,singular perturbations and dimension reduction of multiscale systems.In order to explain the basic ideas,let us consider a two-player stochastic zero-sum dif-ferential game with dynamics given bydX t,ζ;u,vs =bX t,ζ;u,vs,u s,v sds+σX t,ζ;u,v,u s,v sdB s,s∈[t,+∞),X t=ζ,where B is a d-dimensional standard Brownian motion on a given probability space (Ω,F,P),b:R N×U×V→R N andσ:R N×U×V→R N×d,U and V being some metric compact sets.We assume that thefirst player,playing with u,aims at minimizing a running payoff :R N×U×V→R(while the second players,playing with v,maximizes). Then it is known that,under some Isaacs’assumption,the game has a value V T which is the viscosity solution of a second order Hamilton–Jacobi equation of the form−∂t V T(t,x)+Hx,D V T(t,x),D2V T(t,x)=0in[0,T]×R N,V T(T,x)=0in R N.A natural question is the behavior of V T as T→+∞.Actually,since V T is typically of linear growth,the natural quantity to consider is the long-time average,i.e.,lim T→+∞V T/T.Interesting phenomena can be observed under some compactness assumption on the un-derlying state-space.Let us assume,for instance,that the maps b(·,u,v),σ(·,u,v)and (·,u,v)are periodic in all space variables:this actually means that the game takes place in the torus R N/Z N.In this framework,the long-time average is well understood in two cases:either the dif-fusion is strongly nondegenerate:∃ν>0,(σσ∗)(x,u,v)≥νI N∀x,u,v,(where the inequality is understood in the sense of quadratic matrices);orσ≡0and H= H(x,ξ)is coercive:lim|ξ|→+∞H(x,ξ)=+∞uniformly with respect to x.(2) In both cases the quantity V T(x,0)/T uniformly converges to the unique constant¯c forwhich the problem¯c+Hx,Dχ(x),D2χ(x)=0in R Nhas a continuous,periodic solutionχ.In particular,the limit is independent of the initial condition.Such kind of results has been proved by Lions,Papanicoulaou and Varadhan[95] forfirst order equations(i.e.,deterministic differential games).For second order equations, the result has been obtained by Alvarez and Bardi in[3],where the authors combine funda-mental contributions of Evans[61,62]and of Arisawa and Lions[7](see also Alvarez and Bardi[4,5],Bettiol[24],Ghosh and Rao[70]).For deterministic differential games(i.e.,σ≡0),the coercivity condition(2)is not very natural:Indeed,it means that one of the players is much more powerful than the other one. However,very little is known without such a condition.Existing results rely on a specific structure of the game:see for instance Bardi[16],Cardaliaguet[46].The difficulty comes from the fact that,in these cases,the limit may depend upon the initial condition(see also Arisawa and Lions[7],Quincampoix and Renault[104]for related issues in a control set-ting).The existence of a limit for large time differential games is certainly one of the main challenges in differential games theory.4Existence of a Value for Zero-sum Differential Games with State Constraints Differential games with state constraints have been considered since the early theory of differential games:we refer to[23,28,66,69,80]for the computation of the solution for several examples of pursuit.We present here recent trends for obtaining the existence of a value for a rather general class of differential games with constraints.This question had been unsolved during a rather long period due to problems we discuss now.The main conceptual difficulty for considering such zero-sum games lies in the fact that players have to achieve their own goal and to satisfy the state constraint.Indeed,it is not clear to decide which players has to be penalized if the state constraint is violated.For this reason,we only consider a specific class of decoupled games where each player controls independently a part of the dynamics.A second mathematical difficulty comes from the fact that players have to use admissible controls i.e.,controls ensuring the trajectory to fulfilthe state constraint.A byproduct of this problem is the fact that starting from two close initial points it is not obvious tofind two close constrained trajectories.This also affects the regularity of value functions associated with admissible controls:The value functions are,in general,not Lipschitz continuous anymore and,consequently,classical viscosity solutions methods for Hamilton–Jacobi equations may fail.4.1Statement of the ProblemWe consider a differential game where thefirst player playing with u,controls afirst systemy (t)=gy(t),u(t),u(t)∈U,y(t0)=y0∈K U,(3) while the second player,playing with v,controls a second systemz (t)=hz(t),v(t),v(t)∈V,z(t0)=z0∈K V.(4)For every time t,thefirst player has to ensure the state constraint y(t)∈K U while the second player has to respect the state constraint z(t)∈K V for any t∈[t0,T].We denote by x(t)= x[t0,x0;u(·),v(·)](t)=(y[t0,y0;u(·)](t),z[t0,z0;v(·)](t))the solution of the systems(3) and(4)associated with an initial data(t0,x0):=(t0,y0,z0)and with a couple of controls (u(·),v(·)).In the following lines we summarize all the assumptions concerning with the vectorfields of the dynamics:⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩(i)U and V are compact subsets of somefinitedimensional spaces(ii)f:R n×U×V→R n is continuous andLipschitz continuous(with Lipschitz constant M)with respect to x∈R n(iii)uf(x,u,v)andvf(x,u,v)are convex for any x(iv)K U={y∈R l,φU(y)≤0}withφU∈C2(R l;R),∇φU(y)=0ifφU(y)=0(v)K V={z∈R m,φV(z)≤0}withφV∈C2(R m;R),∇φV(z)=0ifφV(z)=0(vi)∀y∈∂K U,∃u∈U such that ∇φU(y),g(y,u) <0(vii)∀z∈∂K V,∃v∈V such that ∇φV(z),h(z,v) <0(5)We need to introduce the notion of admissible controls:∀y0∈K U,∀z0∈K V and∀t0∈[0,T]we defineU(t0,y0):=u(·):[t0,+∞)→U measurable|y[t0,y0;u(·)](t)∈K U∀t≥t0V(t0,z0):=v(·):[t0,+∞)→V measurable|z[t0,z0;v(·)](t)∈K V∀t≥t0.Under assumptions(5),the Viability Theorem(see[9,10])ensures that for all x0= (y0,z0)∈K U×K VU(t0,y0)=∅and V(t0,z0)=∅.Throughout the paper we omit t0in the notations U(t0,y0)and U(t0,y0)whenever t0=0.We now describe two quantitative differential games.Let us start with a game with an integral cost:Bolza Type Differential Game Given a running cost L:[0,T]×R N×U×V→R and afinal costΨ:R N→R,we define the payoff associated to an initial position(t0,x0)= (t0,y0,z0)and to a pair of controls(u,v)∈U(t0,y0)×V(t0,z0)byJt0,x0;u(·),v(·)=Tt0Lt,x(t),u(·),v(·)dt+Ψx(T),(6)where x(t)=x[t0,x0;u(·),v(·)](t)=(y[t0,y0;u(·)](t),z[t0,z0;v(·)](t))denotes the solu-tion of the systems(3)and(4).Thefirst player wants to maximize the functional J,while the second player’s goal is to minimize J.Definition1A mapα:V(t0,z0)→U(t0,y0)is a nonanticipating strategy(for thefirst player and for the point(t0,x0):=(t0,y0,z0)∈R+×K U×K V)if,for anyτ>0,for all controls v1(·)and v2(·)belonging to V(t0,z0),which coincide a.e.on[t0,t0+τ],α(v1(·)) andα(v2(·))coincide almost everywhere on[t0,t0+τ].Nonanticipating strategiesβfor the second player are symmetrically defined.For any point x0∈K U×K V and∀t0∈[0,T]we denote by A(t0,x0)and by B(t0,x0)the sets of the nonanticipating strategies for thefirst and the second player respectively.We are now ready to define the value functions of the game.The lower value V−is defined by:V−(t0,x0):=infβ∈B(t0,x0)supu(·)∈U(t0,y0)Jt0,x0;u(·),βu(·),(7)where J is defined by(6).On the other hand we define the upper value function as follows:V+(t0,x0):=limε→0+supα∈A(t0,x0)infv(·)∈V(t0,z0)Jεt0,x0;αv(·),v(·)(8)withJεt0,x0;u(·),v(·):=Tt0Lt,x(t),u(t),v(t)dt+Ψεx(T),where x(t)=x[t0,x0;u(·),v(·)](t)andΨεis the lower semicontinuous function defined byΨε(x):=infρ∈R|∃y∈R n with(y,ρ)−x,Ψ(x)=ε.The asymmetry between the definition of the value functions is due to the fact that one assumes that the terminal payoffΨis lower semicontinuous.WhenΨis continuous,one can check that V+can equivalently be defined in a more natural way asV+(t0,x0):=supα∈A(t0,x0)infv(·)∈V(t0,z0)Jt0,x0;αv(·),v(·).We now describe the second differential game which is a pursuit game with closed target C⊂K U×K V.Pursuit Type Differential Game The hitting time of C for a trajectory x(·):=(y(·),z(·)) is:θCx(·):=inft≥0|x(t)∈C.If x(t)/∈C for every t≥0,then we setθC(x(·)):=+∞.In the pursuit game,thefirst player wants to maximizeθC while the second player wants to minimize it.The value functions aredefined as follows:The lower optimal hitting-time function is the mapϑ−C :K U×K V→R+∪{+∞}defined,for any x0:=(y0,z0),byϑ−C (x0):=infβ(·)∈B(x0)supu(·)∈U(y0)θCxx0,u(·),βu(·).The upper optimal hitting-time function is the mapϑ+C :K U×K V→R+∪{+∞}de-fined,for any x0:=(y0,z0),byϑ+ C (x0):=limε→0+supα(·)∈A(x0)infv(·)∈V(z0)θC+εBxx0,αv(·),v(·).By convention,we setϑ−C (x)=ϑ+C(x)=0on C.Remarks–Note that here again the definition of the upper and lower value functions are not sym-metric:this is related to the fact that the target assumed to be closed,so that the game is intrinsically asymmetric.–The typical pursuit game is the case when the target coincides with the diagonal:C= {(y,z),|y=z}.We refer the reader to[6,29]for various types of pursuit games.The formalism of the present survey is adapted from[50].4.2Main ResultThe main difficulty for the analysis of state-constraint problems lies in the fact that two trajectories of a control system starting from two—close—different initial conditions could be estimated by classical arguments on the continuity of theflow of the differential equation. For constrained systems,it is easy to imagine cases where the constrained trajectories starting from two close initial conditions are rather far from each other.So,an important problem in order to get suitable estimates on constrained trajectories,is to obtain a kind of Filippov Theorem with ly a result which allows one to approach—in a suitable sense—a given trajectory of the dynamics by a constrained trajectory.Note that similar results exist in the literature.However,we need here to construct a constrained trajectory in a nonanticipating way[26](cf.also[25]),which is not the case in the previous constructions.Proposition1Assume that conditions(5)are satisfied.For any R>0there exist C0= C0(R)>0such that for any initial time t0∈[0,T],for any y0,y1∈K U with|y0|,|y1|≤R,。
Survey of clustering data mining techniques
A Survey of Clustering Data Mining TechniquesPavel BerkhinYahoo!,Inc.pberkhin@Summary.Clustering is the division of data into groups of similar objects.It dis-regards some details in exchange for data simplifirmally,clustering can be viewed as data modeling concisely summarizing the data,and,therefore,it re-lates to many disciplines from statistics to numerical analysis.Clustering plays an important role in a broad range of applications,from information retrieval to CRM. Such applications usually deal with large datasets and many attributes.Exploration of such data is a subject of data mining.This survey concentrates on clustering algorithms from a data mining perspective.1IntroductionThe goal of this survey is to provide a comprehensive review of different clus-tering techniques in data mining.Clustering is a division of data into groups of similar objects.Each group,called a cluster,consists of objects that are similar to one another and dissimilar to objects of other groups.When repre-senting data with fewer clusters necessarily loses certainfine details(akin to lossy data compression),but achieves simplification.It represents many data objects by few clusters,and hence,it models data by its clusters.Data mod-eling puts clustering in a historical perspective rooted in mathematics,sta-tistics,and numerical analysis.From a machine learning perspective clusters correspond to hidden patterns,the search for clusters is unsupervised learn-ing,and the resulting system represents a data concept.Therefore,clustering is unsupervised learning of a hidden data concept.Data mining applications add to a general picture three complications:(a)large databases,(b)many attributes,(c)attributes of different types.This imposes on a data analysis se-vere computational requirements.Data mining applications include scientific data exploration,information retrieval,text mining,spatial databases,Web analysis,CRM,marketing,medical diagnostics,computational biology,and many others.They present real challenges to classic clustering algorithms. These challenges led to the emergence of powerful broadly applicable data2Pavel Berkhinmining clustering methods developed on the foundation of classic techniques.They are subject of this survey.1.1NotationsTo fix the context and clarify terminology,consider a dataset X consisting of data points (i.e.,objects ,instances ,cases ,patterns ,tuples ,transactions )x i =(x i 1,···,x id ),i =1:N ,in attribute space A ,where each component x il ∈A l ,l =1:d ,is a numerical or nominal categorical attribute (i.e.,feature ,variable ,dimension ,component ,field ).For a discussion of attribute data types see [106].Such point-by-attribute data format conceptually corresponds to a N ×d matrix and is used by a majority of algorithms reviewed below.However,data of other formats,such as variable length sequences and heterogeneous data,are not uncommon.The simplest subset in an attribute space is a direct Cartesian product of sub-ranges C = C l ⊂A ,C l ⊂A l ,called a segment (i.e.,cube ,cell ,region ).A unit is an elementary segment whose sub-ranges consist of a single category value,or of a small numerical bin.Describing the numbers of data points per every unit represents an extreme case of clustering,a histogram .This is a very expensive representation,and not a very revealing er driven segmentation is another commonly used practice in data exploration that utilizes expert knowledge regarding the importance of certain sub-domains.Unlike segmentation,clustering is assumed to be automatic,and so it is a machine learning technique.The ultimate goal of clustering is to assign points to a finite system of k subsets (clusters).Usually (but not always)subsets do not intersect,and their union is equal to a full dataset with the possible exception of outliersX =C 1 ··· C k C outliers ,C i C j =0,i =j.1.2Clustering Bibliography at GlanceGeneral references regarding clustering include [110],[205],[116],[131],[63],[72],[165],[119],[75],[141],[107],[91].A very good introduction to contem-porary data mining clustering techniques can be found in the textbook [106].There is a close relationship between clustering and many other fields.Clustering has always been used in statistics [10]and science [158].The clas-sic introduction into pattern recognition framework is given in [64].Typical applications include speech and character recognition.Machine learning clus-tering algorithms were applied to image segmentation and computer vision[117].For statistical approaches to pattern recognition see [56]and [85].Clus-tering can be viewed as a density estimation problem.This is the subject of traditional multivariate statistical estimation [197].Clustering is also widelyA Survey of Clustering Data Mining Techniques3 used for data compression in image processing,which is also known as vec-tor quantization[89].Datafitting in numerical analysis provides still another venue in data modeling[53].This survey’s emphasis is on clustering in data mining.Such clustering is characterized by large datasets with many attributes of different types. Though we do not even try to review particular applications,many important ideas are related to the specificfields.Clustering in data mining was brought to life by intense developments in information retrieval and text mining[52], [206],[58],spatial database applications,for example,GIS or astronomical data,[223],[189],[68],sequence and heterogeneous data analysis[43],Web applications[48],[111],[81],DNA analysis in computational biology[23],and many others.They resulted in a large amount of application-specific devel-opments,but also in some general techniques.These techniques and classic clustering algorithms that relate to them are surveyed below.1.3Plan of Further PresentationClassification of clustering algorithms is neither straightforward,nor canoni-cal.In reality,different classes of algorithms overlap.Traditionally clustering techniques are broadly divided in hierarchical and partitioning.Hierarchical clustering is further subdivided into agglomerative and divisive.The basics of hierarchical clustering include Lance-Williams formula,idea of conceptual clustering,now classic algorithms SLINK,COBWEB,as well as newer algo-rithms CURE and CHAMELEON.We survey these algorithms in the section Hierarchical Clustering.While hierarchical algorithms gradually(dis)assemble points into clusters (as crystals grow),partitioning algorithms learn clusters directly.In doing so they try to discover clusters either by iteratively relocating points between subsets,or by identifying areas heavily populated with data.Algorithms of thefirst kind are called Partitioning Relocation Clustering. They are further classified into probabilistic clustering(EM framework,al-gorithms SNOB,AUTOCLASS,MCLUST),k-medoids methods(algorithms PAM,CLARA,CLARANS,and its extension),and k-means methods(differ-ent schemes,initialization,optimization,harmonic means,extensions).Such methods concentrate on how well pointsfit into their clusters and tend to build clusters of proper convex shapes.Partitioning algorithms of the second type are surveyed in the section Density-Based Partitioning.They attempt to discover dense connected com-ponents of data,which areflexible in terms of their shape.Density-based connectivity is used in the algorithms DBSCAN,OPTICS,DBCLASD,while the algorithm DENCLUE exploits space density functions.These algorithms are less sensitive to outliers and can discover clusters of irregular shape.They usually work with low-dimensional numerical data,known as spatial data. Spatial objects could include not only points,but also geometrically extended objects(algorithm GDBSCAN).4Pavel BerkhinSome algorithms work with data indirectly by constructing summaries of data over the attribute space subsets.They perform space segmentation and then aggregate appropriate segments.We discuss them in the section Grid-Based Methods.They frequently use hierarchical agglomeration as one phase of processing.Algorithms BANG,STING,WaveCluster,and FC are discussed in this section.Grid-based methods are fast and handle outliers well.Grid-based methodology is also used as an intermediate step in many other algorithms (for example,CLIQUE,MAFIA).Categorical data is intimately connected with transactional databases.The concept of a similarity alone is not sufficient for clustering such data.The idea of categorical data co-occurrence comes to the rescue.The algorithms ROCK,SNN,and CACTUS are surveyed in the section Co-Occurrence of Categorical Data.The situation gets even more aggravated with the growth of the number of items involved.To help with this problem the effort is shifted from data clustering to pre-clustering of items or categorical attribute values. Development based on hyper-graph partitioning and the algorithm STIRR exemplify this approach.Many other clustering techniques are developed,primarily in machine learning,that either have theoretical significance,are used traditionally out-side the data mining community,or do notfit in previously outlined categories. The boundary is blurred.In the section Other Developments we discuss the emerging direction of constraint-based clustering,the important researchfield of graph partitioning,and the relationship of clustering to supervised learning, gradient descent,artificial neural networks,and evolutionary methods.Data Mining primarily works with large databases.Clustering large datasets presents scalability problems reviewed in the section Scalability and VLDB Extensions.Here we talk about algorithms like DIGNET,about BIRCH and other data squashing techniques,and about Hoffding or Chernoffbounds.Another trait of real-life data is high dimensionality.Corresponding de-velopments are surveyed in the section Clustering High Dimensional Data. The trouble comes from a decrease in metric separation when the dimension grows.One approach to dimensionality reduction uses attributes transforma-tions(DFT,PCA,wavelets).Another way to address the problem is through subspace clustering(algorithms CLIQUE,MAFIA,ENCLUS,OPTIGRID, PROCLUS,ORCLUS).Still another approach clusters attributes in groups and uses their derived proxies to cluster objects.This double clustering is known as co-clustering.Issues common to different clustering methods are overviewed in the sec-tion General Algorithmic Issues.We talk about assessment of results,de-termination of appropriate number of clusters to build,data preprocessing, proximity measures,and handling of outliers.For reader’s convenience we provide a classification of clustering algorithms closely followed by this survey:•Hierarchical MethodsA Survey of Clustering Data Mining Techniques5Agglomerative AlgorithmsDivisive Algorithms•Partitioning Relocation MethodsProbabilistic ClusteringK-medoids MethodsK-means Methods•Density-Based Partitioning MethodsDensity-Based Connectivity ClusteringDensity Functions Clustering•Grid-Based Methods•Methods Based on Co-Occurrence of Categorical Data•Other Clustering TechniquesConstraint-Based ClusteringGraph PartitioningClustering Algorithms and Supervised LearningClustering Algorithms in Machine Learning•Scalable Clustering Algorithms•Algorithms For High Dimensional DataSubspace ClusteringCo-Clustering Techniques1.4Important IssuesThe properties of clustering algorithms we are primarily concerned with in data mining include:•Type of attributes algorithm can handle•Scalability to large datasets•Ability to work with high dimensional data•Ability tofind clusters of irregular shape•Handling outliers•Time complexity(we frequently simply use the term complexity)•Data order dependency•Labeling or assignment(hard or strict vs.soft or fuzzy)•Reliance on a priori knowledge and user defined parameters •Interpretability of resultsRealistically,with every algorithm we discuss only some of these properties. The list is in no way exhaustive.For example,as appropriate,we also discuss algorithms ability to work in pre-defined memory buffer,to restart,and to provide an intermediate solution.6Pavel Berkhin2Hierarchical ClusteringHierarchical clustering builds a cluster hierarchy or a tree of clusters,also known as a dendrogram.Every cluster node contains child clusters;sibling clusters partition the points covered by their common parent.Such an ap-proach allows exploring data on different levels of granularity.Hierarchical clustering methods are categorized into agglomerative(bottom-up)and divi-sive(top-down)[116],[131].An agglomerative clustering starts with one-point (singleton)clusters and recursively merges two or more of the most similar clusters.A divisive clustering starts with a single cluster containing all data points and recursively splits the most appropriate cluster.The process contin-ues until a stopping criterion(frequently,the requested number k of clusters) is achieved.Advantages of hierarchical clustering include:•Flexibility regarding the level of granularity•Ease of handling any form of similarity or distance•Applicability to any attribute typesDisadvantages of hierarchical clustering are related to:•Vagueness of termination criteria•Most hierarchical algorithms do not revisit(intermediate)clusters once constructed.The classic approaches to hierarchical clustering are presented in the sub-section Linkage Metrics.Hierarchical clustering based on linkage metrics re-sults in clusters of proper(convex)shapes.Active contemporary efforts to build cluster systems that incorporate our intuitive concept of clusters as con-nected components of arbitrary shape,including the algorithms CURE and CHAMELEON,are surveyed in the subsection Hierarchical Clusters of Arbi-trary Shapes.Divisive techniques based on binary taxonomies are presented in the subsection Binary Divisive Partitioning.The subsection Other Devel-opments contains information related to incremental learning,model-based clustering,and cluster refinement.In hierarchical clustering our regular point-by-attribute data representa-tion frequently is of secondary importance.Instead,hierarchical clustering frequently deals with the N×N matrix of distances(dissimilarities)or sim-ilarities between training points sometimes called a connectivity matrix.So-called linkage metrics are constructed from elements of this matrix.The re-quirement of keeping a connectivity matrix in memory is unrealistic.To relax this limitation different techniques are used to sparsify(introduce zeros into) the connectivity matrix.This can be done by omitting entries smaller than a certain threshold,by using only a certain subset of data representatives,or by keeping with each point only a certain number of its nearest neighbors(for nearest neighbor chains see[177]).Notice that the way we process the original (dis)similarity matrix and construct a linkage metric reflects our a priori ideas about the data model.A Survey of Clustering Data Mining Techniques7With the(sparsified)connectivity matrix we can associate the weighted connectivity graph G(X,E)whose vertices X are data points,and edges E and their weights are defined by the connectivity matrix.This establishes a connection between hierarchical clustering and graph partitioning.One of the most striking developments in hierarchical clustering is the algorithm BIRCH.It is discussed in the section Scalable VLDB Extensions.Hierarchical clustering initializes a cluster system as a set of singleton clusters(agglomerative case)or a single cluster of all points(divisive case) and proceeds iteratively merging or splitting the most appropriate cluster(s) until the stopping criterion is achieved.The appropriateness of a cluster(s) for merging or splitting depends on the(dis)similarity of cluster(s)elements. This reflects a general presumption that clusters consist of similar points.An important example of dissimilarity between two points is the distance between them.To merge or split subsets of points rather than individual points,the dis-tance between individual points has to be generalized to the distance between subsets.Such a derived proximity measure is called a linkage metric.The type of a linkage metric significantly affects hierarchical algorithms,because it re-flects a particular concept of closeness and connectivity.Major inter-cluster linkage metrics[171],[177]include single link,average link,and complete link. The underlying dissimilarity measure(usually,distance)is computed for every pair of nodes with one node in thefirst set and another node in the second set.A specific operation such as minimum(single link),average(average link),or maximum(complete link)is applied to pair-wise dissimilarity measures:d(C1,C2)=Op{d(x,y),x∈C1,y∈C2}Early examples include the algorithm SLINK[199],which implements single link(Op=min),Voorhees’method[215],which implements average link (Op=Avr),and the algorithm CLINK[55],which implements complete link (Op=max).It is related to the problem offinding the Euclidean minimal spanning tree[224]and has O(N2)complexity.The methods using inter-cluster distances defined in terms of pairs of nodes(one in each respective cluster)are called graph methods.They do not use any cluster representation other than a set of points.This name naturally relates to the connectivity graph G(X,E)introduced above,because every data partition corresponds to a graph partition.Such methods can be augmented by so-called geometric methods in which a cluster is represented by its central point.Under the assumption of numerical attributes,the center point is defined as a centroid or an average of two cluster centroids subject to agglomeration.It results in centroid,median,and minimum variance linkage metrics.All of the above linkage metrics can be derived from the Lance-Williams updating formula[145],d(C iC j,C k)=a(i)d(C i,C k)+a(j)d(C j,C k)+b·d(C i,C j)+c|d(C i,C k)−d(C j,C k)|.8Pavel BerkhinHere a,b,c are coefficients corresponding to a particular linkage.This formula expresses a linkage metric between a union of the two clusters and the third cluster in terms of underlying nodes.The Lance-Williams formula is crucial to making the dis(similarity)computations feasible.Surveys of linkage metrics can be found in [170][54].When distance is used as a base measure,linkage metrics capture inter-cluster proximity.However,a similarity-based view that results in intra-cluster connectivity considerations is also used,for example,in the original average link agglomeration (Group-Average Method)[116].Under reasonable assumptions,such as reducibility condition (graph meth-ods satisfy this condition),linkage metrics methods suffer from O N 2 time complexity [177].Despite the unfavorable time complexity,these algorithms are widely used.As an example,the algorithm AGNES (AGlomerative NESt-ing)[131]is used in S-Plus.When the connectivity N ×N matrix is sparsified,graph methods directly dealing with the connectivity graph G can be used.In particular,hierarchical divisive MST (Minimum Spanning Tree)algorithm is based on graph parti-tioning [116].2.1Hierarchical Clusters of Arbitrary ShapesFor spatial data,linkage metrics based on Euclidean distance naturally gener-ate clusters of convex shapes.Meanwhile,visual inspection of spatial images frequently discovers clusters with curvy appearance.Guha et al.[99]introduced the hierarchical agglomerative clustering algo-rithm CURE (Clustering Using REpresentatives).This algorithm has a num-ber of novel features of general importance.It takes special steps to handle outliers and to provide labeling in assignment stage.It also uses two techniques to achieve scalability:data sampling (section 8),and data partitioning.CURE creates p partitions,so that fine granularity clusters are constructed in parti-tions first.A major feature of CURE is that it represents a cluster by a fixed number,c ,of points scattered around it.The distance between two clusters used in the agglomerative process is the minimum of distances between two scattered representatives.Therefore,CURE takes a middle approach between the graph (all-points)methods and the geometric (one centroid)methods.Single and average link closeness are replaced by representatives’aggregate closeness.Selecting representatives scattered around a cluster makes it pos-sible to cover non-spherical shapes.As before,agglomeration continues until the requested number k of clusters is achieved.CURE employs one additional trick:originally selected scattered points are shrunk to the geometric centroid of the cluster by a user-specified factor α.Shrinkage suppresses the affect of outliers;outliers happen to be located further from the cluster centroid than the other scattered representatives.CURE is capable of finding clusters of different shapes and sizes,and it is insensitive to outliers.Because CURE uses sampling,estimation of its complexity is not straightforward.For low-dimensional data authors provide a complexity estimate of O (N 2sample )definedA Survey of Clustering Data Mining Techniques9 in terms of a sample size.More exact bounds depend on input parameters: shrink factorα,number of representative points c,number of partitions p,and a sample size.Figure1(a)illustrates agglomeration in CURE.Three clusters, each with three representatives,are shown before and after the merge and shrinkage.Two closest representatives are connected.While the algorithm CURE works with numerical attributes(particularly low dimensional spatial data),the algorithm ROCK developed by the same researchers[100]targets hierarchical agglomerative clustering for categorical attributes.It is reviewed in the section Co-Occurrence of Categorical Data.The hierarchical agglomerative algorithm CHAMELEON[127]uses the connectivity graph G corresponding to the K-nearest neighbor model spar-sification of the connectivity matrix:the edges of K most similar points to any given point are preserved,the rest are pruned.CHAMELEON has two stages.In thefirst stage small tight clusters are built to ignite the second stage.This involves a graph partitioning[129].In the second stage agglomer-ative process is performed.It utilizes measures of relative inter-connectivity RI(C i,C j)and relative closeness RC(C i,C j);both are locally normalized by internal interconnectivity and closeness of clusters C i and C j.In this sense the modeling is dynamic:it depends on data locally.Normalization involves certain non-obvious graph operations[129].CHAMELEON relies heavily on graph partitioning implemented in the library HMETIS(see the section6). Agglomerative process depends on user provided thresholds.A decision to merge is made based on the combinationRI(C i,C j)·RC(C i,C j)αof local measures.The algorithm does not depend on assumptions about the data model.It has been proven tofind clusters of different shapes,densities, and sizes in2D(two-dimensional)space.It has a complexity of O(Nm+ Nlog(N)+m2log(m),where m is the number of sub-clusters built during the first initialization phase.Figure1(b)(analogous to the one in[127])clarifies the difference with CURE.It presents a choice of four clusters(a)-(d)for a merge.While CURE would merge clusters(a)and(b),CHAMELEON makes intuitively better choice of merging(c)and(d).2.2Binary Divisive PartitioningIn linguistics,information retrieval,and document clustering applications bi-nary taxonomies are very useful.Linear algebra methods,based on singular value decomposition(SVD)are used for this purpose in collaborativefilter-ing and information retrieval[26].Application of SVD to hierarchical divisive clustering of document collections resulted in the PDDP(Principal Direction Divisive Partitioning)algorithm[31].In our notations,object x is a docu-ment,l th attribute corresponds to a word(index term),and a matrix X entry x il is a measure(e.g.TF-IDF)of l-term frequency in a document x.PDDP constructs SVD decomposition of the matrix10Pavel Berkhin(a)Algorithm CURE (b)Algorithm CHAMELEONFig.1.Agglomeration in Clusters of Arbitrary Shapes(X −e ¯x ),¯x =1Ni =1:N x i ,e =(1,...,1)T .This algorithm bisects data in Euclidean space by a hyperplane that passes through data centroid orthogonal to the eigenvector with the largest singular value.A k -way split is also possible if the k largest singular values are consid-ered.Bisecting is a good way to categorize documents and it yields a binary tree.When k -means (2-means)is used for bisecting,the dividing hyperplane is orthogonal to the line connecting the two centroids.The comparative study of SVD vs.k -means approaches [191]can be used for further references.Hier-archical divisive bisecting k -means was proven [206]to be preferable to PDDP for document clustering.While PDDP or 2-means are concerned with how to split a cluster,the problem of which cluster to split is also important.Simple strategies are:(1)split each node at a given level,(2)split the cluster with highest cardinality,and,(3)split the cluster with the largest intra-cluster variance.All three strategies have problems.For a more detailed analysis of this subject and better strategies,see [192].2.3Other DevelopmentsOne of early agglomerative clustering algorithms,Ward’s method [222],is based not on linkage metric,but on an objective function used in k -means.The merger decision is viewed in terms of its effect on the objective function.The popular hierarchical clustering algorithm for categorical data COB-WEB [77]has two very important qualities.First,it utilizes incremental learn-ing.Instead of following divisive or agglomerative approaches,it dynamically builds a dendrogram by processing one data point at a time.Second,COB-WEB is an example of conceptual or model-based learning.This means that each cluster is considered as a model that can be described intrinsically,rather than as a collection of points assigned to it.COBWEB’s dendrogram is calleda classification tree.Each tree node(cluster)C is associated with the condi-tional probabilities for categorical attribute-values pairs,P r(x l=νlp|C),l=1:d,p=1:|A l|.This easily can be recognized as a C-specific Na¨ıve Bayes classifier.During the classification tree construction,every new point is descended along the tree and the tree is potentially updated(by an insert/split/merge/create op-eration).Decisions are based on the category utility[49]CU{C1,...,C k}=1j=1:kCU(C j)CU(C j)=l,p(P r(x l=νlp|C j)2−(P r(x l=νlp)2.Category utility is similar to the GINI index.It rewards clusters C j for in-creases in predictability of the categorical attribute valuesνlp.Being incre-mental,COBWEB is fast with a complexity of O(tN),though it depends non-linearly on tree characteristics packed into a constant t.There is a similar incremental hierarchical algorithm for all numerical attributes called CLAS-SIT[88].CLASSIT associates normal distributions with cluster nodes.Both algorithms can result in highly unbalanced trees.Chiu et al.[47]proposed another conceptual or model-based approach to hierarchical clustering.This development contains several different use-ful features,such as the extension of scalability preprocessing to categori-cal attributes,outliers handling,and a two-step strategy for monitoring the number of clusters including BIC(defined below).A model associated with a cluster covers both numerical and categorical attributes and constitutes a blend of Gaussian and multinomial models.Denote corresponding multivari-ate parameters byθ.With every cluster C we associate a logarithm of its (classification)likelihoodl C=x i∈Clog(p(x i|θ))The algorithm uses maximum likelihood estimates for parameterθ.The dis-tance between two clusters is defined(instead of linkage metric)as a decrease in log-likelihoodd(C1,C2)=l C1+l C2−l C1∪C2caused by merging of the two clusters under consideration.The agglomerative process continues until the stopping criterion is satisfied.As such,determina-tion of the best k is automatic.This algorithm has the commercial implemen-tation(in SPSS Clementine).The complexity of the algorithm is linear in N for the summarization phase.Traditional hierarchical clustering does not change points membership in once assigned clusters due to its greedy approach:after a merge or a split is selected it is not refined.Though COBWEB does reconsider its decisions,its。
Suwon University, Gyeonggi-do, Korea SPONSORED BY
FINAL PROGRAMTHE 2007 ACM SIGAPPSYMPOSIUM ON APPLIED COMPUTING/conferences/sac/sac2007Seoul, Korea March 11 - 15, 2007Organizing CommitteeRoger L. Wainwright Hisham M. Haddad Sung Y. ShinSascha Ossowski Ronaldo MenezesLorie M. Liebrock Mathew J. Palakal Jaeyoung Choi Tei-Wei Kuo Jiman HongSeong Tae Jhang Yookun Cho Yong Wan KooH OSTED BYSeoul National University, Seoul, Korea Suwon University, Gyeonggi-do, KoreaSPONSORED BYSAC 2007 I NTRODUCTIONSAC 2007 is a premier international conference on applied com-puting and technology. Attendees have the opportunity to hear from expert practitioners and researchers about the latest trends in research and development in their fields. SAC 2007 features 2 keynote speakers on Monday and Wednesday, from 8:30 to 10:00. The symposium consists of Tutorial and Technical programs. The Tutorial Program offers 3 half-day tutorials on Sunday March 11, 2007, starting at 9:00am. The Technical Program offers 38 tracks on a wide number of different research topics, which run from Monday March 12 through Thursday March 15, 2007. Regular sessions start at 8:30am and end at 5:00pm in 4 parallel sessions. Honorable ChairsYookun Cho, Honorable Symposium ChairSeoul National University, KoreaYong Wan Koo, Honorable Program ChairUniversity of Suwon, KoreaOrganizing CommitteeRoger L. Wainwright, Symposium ChairUniversity of Tulsa, USAHisham M. Haddad, Symposium Chair, Treasurer, Registrar Kennesaw State University, USASung Y. Shin, Symposium ChairSouth Dakota State University, USASascha Ossowski, Program ChairUniversity Rey Juan Carlos, Madrid, SpainRonaldo Menezes, Program ChairFlorida Institute of Technology, Melbourne, FloridaJaeyoung Choi, Tutorials ChairSoongsil University, KoreaTei-Wei Kuo, Tutorials ChairNational Taiwan University, ChinaMathew J. Palakal, Poster ChairIndiana University Purdue University, USALorie M. Liebrock, Publication ChairNew Mexico Institute of Mining and Technology, USAJiman Hong,Local Organization ChairKwangwoon University, KoreaSeong Tae Jhang,Local Organization ChairUniversity of Suwon, KoreaSAC 2007 Track OrganizersArtificial Intelligence, Computational Logic, and Image Analysis (AI)C.C. Hung, School of Computing and Soft. Eng., USAAgostinho Rosa, LaSEEB –ISR – IST, PortugalAdvances in Spatial and Image-based Information Systems (ASIIS)Kokou Yetongnon, Bourgogne University, FranceChristophe Claramunt, Naval Academy Research Institute, France Richard Chbeir, Bourgogne University, FranceKi-Joune Li, Prusan National University, KoreaAgents, Interactions, Mobility and Systems (AIMS)Marcin Paprzycki, SWPS and IBS PAN, PolandCostin Badica, University of Craiova, RomaniaMaria Ganzha, EUH-E and IBS PAN, PolandAlex Yung-Chuan Lee, Southern Illinois University, USAShahram Rahimi, Southern Illinois University, USAAutonomic Computing (AC)Umesh Bellur, Indian Institute of Technology, IndiaSheikh Iqbal Ahamed, Marquette University, USABioinformatics (BIO)Mathew J. Palakal, Indiana University Purdue University, USALi Liao, University of Delaware, USAComputer Applications in Health Care (CACH)Valentin Masero, University of Extremadura, SpainPierre Collet, Université du Littoral (ULCO), France Computer Ethics and Human Values (CEHV)Kenneth E. Himma, Seattle Pacific University, USAKeith W. Miller, University of Illinois at Springfield, USADavid S. Preston, University of East London, UKComputer Forensics (CF)Brajendra Panda, University of Arkansas, USAKamesh Namuduri, Wichita State University, USAComputer Networks (CN)Mario Freire, University of Beira Interior, PortugalTeresa Vazao, INESC ID/IST, PortugalEdmundo Monteiro, University of Coimbra, PortugalManuela Pereira, University of Beira Interior, PortugalComputer Security (SEC)Giampaolo Bella, Universita' di Catania, ItalyPeter Ryan, University of Newcastle upon Tyne, UKComputer-aided Law and Advanced Technologies (CLAT) Giovanni Sartor, University of Bologna, ItalyAlessandra Villecco Bettelli, University of Bologna, ItalyLavinia Egidi, University of Piemonte Orientale, ItalyConstraint Solving and Programming (CSP)Stefano Bistarelli, Università degli studi "G. D'Annunzio" di Chieti-Pescara, ItalyEric Monfroy, University of Nantes, FranceBarry O'Sullivan, University College Cork, IrelandCoordination Models, Languages and Applications (CM) Alessandro Ricci, Universita di Bologna, ItalyBernhard Angerer, Michael Ignaz Schumacher, EPFL IC IIF LIA, SwitzerlandData Mining (DM)Hasan M. Jamil, Wayne State University, USAData Streams (DS)Jesus S. Aguilar-Ruiz, Pablo de Olavide University, SpainFrancisco J. Ferrer-Troyano, University of Seville, SpainJoao Gama, University of Porto, PortugalRalf Klinkenberg, University of Dortmund, GermanyDatabase Theory, Technology, and Applications (DTTA) Ramzi A. Haraty, Lebanese American University, LebanonApostolos N. Papadopoulos, Aristotle University, GreeceJunping Sun, Nova Southeastern University, USADependable and Adaptive Distributed Systems (DADS)Karl M. Göschka, Vienna University of Technology, AustriaSvein O. Hallsteinsen, SINTEF ICT, NorwayRui Oliveira, Universidade do Minho, PortugalAlexander Romanovsky, University of Newcastle upon Tyne, UK Document Engineering (DE)Rafael Dueire Lins, Universidade Federal de Pernambuco, Brazil Electronic Commerce Technologies (ECT)Sviatoslav Braynov, University of Illinois at Springfield, USADaryl Nord, Oklahoma State University, USAFernando Rubio, Universidad Complutense de Madrid, Spain Embedded Systems: Applications, Solutions and Techniques (EMBS)Alessio Bechini, University of Pisa, ItalyCosimo Antonio Prete, University of Pisa, ItalyJihong Kim, Seoul National University, KoreaEvolutionary Computation (EC)Bryant A. Julstrom, St. Cloud State University, USA Geoinformatics and Technology (GT)Dong-Cheon Lee, Sejong University, KoreaGwangil Jeon, Korea Polytechnic University, KoreaGeometric Computing and Reasoning (GCR)Xiao-Shan Gao, Chinese Academy of Sciences, ChinaDominique Michelucci, Universite de Bourgogne, FrancePascal Schreck, Universite Louis Pasteur, FranceHandheld Computing (HHC)Qusay H. Mahmoud, University of Guelph, CanadaZakaria Maamar, Zayed University, UAEInformation Access and Retrieval (IAR)Fabio Crestani, University of Strathclyde, UKGabriella Pasi, University of Milano Bicocca, ItalyMobile Computing and Applications (MCA)Hong Va Leong, Hong Kong Polytechnic University, Hong KongAlvin Chan, Hong Kong Polytechnic University, Hong KongModel Transformation (MT)Jean Bézivin, University of Nantes, FranceAlfonso Pierantonio, Università degli Studi dell’Aquila, ItalyAntonio Vallecillo, Universidad de Malaga, SpainJeff Gray, University of Alabama at Birmingham, USAMultimedia and Visualization (MMV)Chaman L. Sabharwal, University of Missouri-Rolla, USAMingjun Zhang, Agilent Technologies, USAObject-Oriented Programming Languages and Systems (OOP) Davide Ancona, DISI - Università di Genova, ItalyMirko Viroli, Università di Bologna, ItalyOperating Systems and Adaptive Applications (OSAA)Jiman Hong, Kwangwoon University, KoreaTei-Wei Kuo, National Taiwan University, TaiwanOrganizational Engineering (OE)José Tribolet, Technical University of Lisbon, PortugalRobert Winter, University of St. Gallen, SwitzerlandArtur Caetano, Technical University of Lisbon, Portugal Programming for Separation of Concerns (PSC)Corrado Santoro, Catania University, ItalyEmiliano Tramontana, Catania University, ItalyIan Welch, Victoria University, New ZealandYvonne Coady, Victoria Univeristy, CanadaProgramming Languages (PL)Chang-Hyun Jo, California State University at Fullerton, USAMarjan Mernik, University of Maribor, SloveniaBarrett Bryant, University of Alabama at Birmingham, USAReliable Computations and their Applications (RCA)Martine Ceberio, University of Texas at El Paso, USAVladik Kreinovich, University of Texas at El Paso, USAMichael Rueher, Universite de Nice ESSI, FranceSemantic Web and Application (SWA)Hyoil Han, Drexel University, USASemantic-Based Resource Discovery, Retrieval and Composition (SDRC)Eugenio Di Sciascio, SinsInfLab Politecnico di Bari, ItalyFrancesco M. Donini, University of Tuscia, ItalyTommaso Di Noia, SinsInfLab Politecnico di Bari, ItalyMassimo Paolucci, DoCoMo Euro-Labs, GermanySoftware Engineering (SE)W. Eric Wong, University of Texas at Dallas, USAChang-Oan Sung, Indiana University Southeast, USASoftware Verification (SV)Zijiang Yang, Western Michigan University, USALunjin Lu, Oakland University, USAFausto Spoto, Universita di Verona, ItalySystem On Chip Design and Software Supports (SODSS) Seong Tae Jhang, Suwon University, KoreaSung Woo Chung, Korea University, KoreaTrust, Recommendations, Evidence and other Collaborative Know-how (TRECK)Jean-Marc Seigneur, University of Geneva, SwitzerlandJeong Hyun Yi, Samsung Advanced Institute of Technology, South Korea Ubiquitous Computing: Digital Spaces, Services and Content (UC)Achilles Kameas, Hellenic Open University, GreeceGeorge Roussos, University of London, UKWeb Technologies (WT)Fahim Akhter , Zayed University, UAEDjamal Benslimane, University of Lyon, FranceZakaria Maamar, Zayed University, UAEQusay H. Mahmoud, University of Guelph, CanadaLocal SupportLocal support for SAC 2007 is provided by the Seoul National University in Seoul, Suwon University in Gyeonggi-do, Ministry of Education and Human Resources Development, Samsung, mds technology, KETI, MIC, CVB, and ETRI. The SAC organizing committee acknowledges and thanks the local supporters for their generous contributions to SAC 2007. Their support has been essential to the success of Symposium, and is greatly appreciated. ACM SIGAPPThe ACM Special Interest Group on Applied Computing is ACM's primary applications-oriented SIG. Its mission is to further the interests of the computing professionals engaged in the development of new computing applications and applications areas and the transfer of computing technology to new problem domains. SIGAPP offers practitioners and researchers the opportunity to share mutual interests in innovative application fields, technology transfer, experimental computing, strategic research, and the management of computing. SIGAPP also promotes widespread cooperation among business, government, and academic computing activities. Its annual Symposium on Applied Computing (SAC) provides an international forum for presentation of the results of strategic research and experimentation for this inter-disciplinary environment. SIGAPP membership fees are: $30.00 for ACM Non-members, $15.00 for ACM Members, and $8.00 for Student Members. For information contact Barrett Bryant at bryant@. Also, checkout the SIGAPP website at /sigapp/M ESSAGE FROM THE S YMPOSIUM C HAIRSRoger WaiwrightUniversity of Tulsa, USAHisham M. HaddadKennesaw State University, USASung Y. ShinSouth Dakota State University, USAOn behalf of the Organization Committee, it is our pleasure to welcome you to the 22nd Annual ACM Symposium on Applied Computing (SAC 2007). This year, the conference is hosted by Seoul National University and Suwon University in Gyeonggi-do, Korea. Many thanks for your participation in this international event dedicated to computer scientists, engineers, and practitioners seeking innovative ideas in various areas of computer applications. The sponsoring SIG of this Symposium, the ACM Special Interest Group on Applied Computing, is dedicated to further the interests of computing professionals engaged in the design and development of new computing applications, interdisciplinary applications areas, and applied research. The conference provides a forum for discussion and exchange of new ideas addressing computational algorithms and complex applications. This goal is reflected in its wide spectrum of application areas and tutorials designed to provide variety of discussion topics during this event. The conference is composed of various specialized technical tracks and tutorials. As in past successful meetings, talented and dedicated Track Chairs and Co-Chairs have organized SAC 2007 tracks. Each track maintains a program committee and group of highly qualified reviewers. We thank the Track Chairs, Co-Chairs, and participating reviewers for their commitment to making SAC 2007 another high quality conference. We also thank our invited keynote speakers for sharing their knowledge with SAC attendees. Most of all, special thanks to the authors and presenters for sharing their experience with the rest of us and to all attendees for joining us in Seoul, Korea.The local organizing committee has always been a key to the success of the conference. This year, we thank our local team from Seoul National University and Suwon University. In particular, we thank Dr. Jiman Hong, from Kwangwoon University, and Dr. Seong Tae Jhang, from Suwon University, for chairing the local organization effort. We also thank Dr. Jaeyoung Choi, from Soongsil University, and Dr. Tei-Wei Kuo, from National Taiwan University, for organizing the Tutorials Program. Other committee members we also would like to thank are Lorie Liebrock for her tremendous effort putting together the conference proceedings, Mathew Palakal for coordinating another successful Posters Program, and Sascha Ossowski and Ronaldo Menezes for bringing together the Technical Program. Finally, we extend outthanks and gratitude to our honorable Symposium and Program Chairs Drs. Yookun Cho of Seoul National University and Dr. Yong Wan Koo of Suwon University. Many thanks for hosting the conference and coordinating governmental and local support. Again, we welcome you to SAC 2007 in the lively city of Seoul. We hope you enjoy your stay in Seoul and leave this event enriched with new ideas and friends. Next year, we invite you to participate in SAC 2008 to be held in the costal city of Fortaleza, Brazil. The symposium will be hosted by the University of Fortaleza (UNIFOR) and the Federal University of Ceará (UFC). We hope to see there!M ESSAGE FROM THE P ROGRAM C HAIRSSascha OssowskiUniversity Rey Juan Carlos, SpainRonaldo MenezesFlorida Institute of Technology, USAWelcome to the 22nd Symposium on Applied Computing (SAC 2007). Over the past 21 years, SAC has been an international forum for researchers and practitioners to present their findings and research results in the areas of computer applications and technology. The SAC 2007 Technical Program offers a wide range of tracks covering major areas of computer applications. Highly qualified referees with strong expertise and special interest in their respective research areas carefully reviewed the submitted papers. As part of the Technical Program, this year the Tutorial Program offers several half-day tutorials that were carefully selected from numerous proposals. Many thanks to Jaeyoung Choi from the Soongsil University and Tei-Wei Kuo from the National Taiwan University for chairing the Tutorial Program. Also, this is the fourth year for SAC to incorporate poster papers into the Technical Program. Many thanks to Mathew Palakal from Indiana University Purdue University for chairing the poster sessions. SAC 2007 would not be possible without contributions from members of the scientific community. As anyone can imagine, many people have dedicated tremendous time and effort over the period of 10 months to bring you an excellent program. The success of SAC 2007 relies on the effort and hard work of many volunteers. On behalf of the SAC 2007 Organizing Committee, we would like to take this opportunity to thank all of those who made this year's technical program a reality, including speakers, referees, track chairs, session chairs, presenters, and attendees. We also thank the local arrangement committee lead by Jiman Hong from the Kwangwoon University and Seong Tae Jhang from Suwon University. We also want to thank Hisham Haddad from Kennesaw State University for his excellent job again as the SAC Treasurer, Webmaster, and Registrar.SAC's open call for Track Proposals resulted in the submission of 47 track proposals. These proposals were carefully evaluated by the conference Executive Committee. Some proposals were rejected on the grounds of either not being appropriate for the areas that SAC covers traditionally or being of rather narrow and specialized nature. Some others tracks were merged to form a single track. Eventually, 38 tracks were established, which then went on to produce their own call for papers. In response to these calls, 786 papers were submitted, from which 256 papers were strongly recommended by the referees for acceptance and inclusion in the Conference Proceedings. This gives SAC 2007 an acceptance rate of 32.5% across all tracks. SAC is today one of the most popular and competitive conferences in the international field of applied computing.We hope you will enjoy the meeting and have the opportunity to exchange your ideas and make new friends. We also hope you will enjoy your stay in Seoul, Korea and take pleasure from the many entertainments and activities that the city and Korea has to offer. We look forward to your active participation in SAC 2008 when for the first time SAC will be hosted in South America, more specifically in Fortaleza, Brazil. We encourage you and your colleagues to submit your research findings to next year's technical program. Thank you for being part of SAC 2007, and we hope to see you in sunny Fortaleza, Brazil for SAC 2008.O THER A CTIVITIESReview Meeting: Sunday March 11, 2007, from 18:00 to 19:00 in Room 311A. Open for SAC Organizing Committee and Track Chairs and Co-Chairs.SAC 2008 Organization Meeting: Monday March 12, 2007, from 18:00 to 19:00 in Room 311A. Open for SAC Organizing Committee.SAC Reception: Monday March 12, 2007 at 19:00 to 22:00. Room 402. Open for all registered attendees.Posters Session: Tuesday March 13, 2007, from 13:30 to 17:00 in the Room 311C. Open to everyone.SIGAPP Annual Business Meeting: Tuesday March 13, 2007, from 17:15 to 18:15 in Room 311A. Open to everyone.SAC Banquet: Wednesday March 14, 2007. Rooms 331-334. Open for Banquet Ticket holders. See your tickets for full details. Track-Chairs Luncheon: Thursday April 27, 2006, from 12:00 to 13:30. Hosu (Lake) Food-mall. Open for SAC Organizing Committee, Track Chairs and Co-Chairs.SAC 2008SAC 2008 will be held in Fortaleza, Ceará, Brazil, March 16 – 20, 2008. It is co-hosted by the University of Fortaleza (UNIFOR) and the Federal University of Ceará (UFC). Please check the registration desk for handouts. You can also visit the website at /conferences/sac/sac2008/.M ONDAY K EYNOTE A DDRESSA New DBMS Architecture for DB-IRIntegrationDr. Kyu-Young WhangDirector of Advanced Information Technology Research Center, Korea Advanced Institute ofScience and Technology, Daejeon, Korea M ONDAY M ARCH 12, 2007, 9:00 – 10:00ROOM 310 A, B AND CABSTRACTNowadays, there is an increasing need to integrate the DBMS (for structured data) with Information Retrieval (IR) features (for unstructured data). DB-IR integration becomes one of major challenges in the database area. Extensible architectures provided by commercial ORDBMS vendors can be used for DB-IR integration. Here, extensions are implemented using a high-level (typically, SQL-level) interface. We call this architecture loose-coupling. The advantage of loose-coupling is that it is easy to implement. But, it is not preferable for implementing new data types and operations in large databases when high performance is required. In this talk, we present a new DBMS architectureapplicable to DB-IR integration, which we call tight-coupling. In tight-coupling, new data types and operations are integrated into the core of the DBMS engine in the extensible type layer. Thus, they are incorporated as the "first-class citizens" within the DBMS architecture and are supported in a consistent manner with high performance. This tight-coupling architecture is being used to incorporate IR features and spatial database features into the Odysseus ORDBMS that has been under development at KAIST/AITrc for over 16 years. In this talk, we introduce Odysseus and explain its tightly-coupled IR features (U.S. patented in 2002). Then, we demonstrate excellence of tight-coupling by showing benchmark results. We have built a web search engine that is capable of managing 20~100 million web pages in a non-parallel configuration using Odysseus. This engine has been successfully tested in many commercial environments. In a parallel configuration, it is capable of managing billons of web pages. This work won the Best Demonstration Award from the IEEE ICDE conference held in Tokyo, Japan in April 2005.W EDNESDAY K EYNOTE A DDRESS The Evolution of Digital Evidence asa Forensic ScienceDr. Marc RogersChair of the Cyber Forensics Program,Department of Computer and InformationTechnology, Purdue University, USAW EDNESDAY M ARCH 14, 2007, 9:00 –10:00ROOMS 310 A, B AND CABSTRACTThe field of Digital Evidence while garnering significant attention by academia, the public, and the media, has really just begun its journey as a forensic science. Digital Forensic Science (DFS) in general is an immature discipline in comparison to the other more traditional forensic sciences such as latent fingerprint analysis. Digital Evidence, which falls under the larger umbrella of DFS, truly encompasses the notion of being an applied multi-disciplinary science. The areas of Computer Science, Technology, Engineering, Mathematics, Law, Sociology, Psychology, Criminal Justice etc. all have played and will continue to play a very large role in maturing and defining this scientific field. The presentation will look at the history of Digital Forensic Science and Digital Evidence, the current state of the field, and what might be in store for the future.S EOUL R EPRESENTATIVE A DDRESSKoran IT policy - IT839Dr. Jung-hee SongAssistant MayorChief of Information OfficerInformation System Planning DivisionSeoul Metropolitan Government, KoreaW EDNESDAY M ARCH 14, 2007, 18:30 – 19:00ROOMS 331-334(DURING BANQUET)ABSTRACTKorean IT policy initiated by Ministry of Information and Communication called IT839 Strategy will be introduced. By defining government role in the u-Korea vision pursuit, it removes uncertainties for IT industry and increases its active participation. As capital of Korea, Seoul presented a grand plan to be u-Seoul. An overview of u-Seoul masterplan will be delivered with introduction of 5 specific projects.SAC 2007 S CHEDULES UNDAY M ARCH 11, 200709:00 – 17:00 L OBBYR EGISTRATION09:00 – 10:30 R OOMS 310 A AND BAM T UTORIALS IT1: Introduction to Security-enhanced Linux(SELinux)Dr. Haklin Kimm, Professor, omputer Science Department, ast Stroudsburg University of Pennsylvania, USAT2: Similarity Search - The Metric Space Approach Pavel Zezula, Masaryk University, Brno, Czech RepublicGiuseppe Amato, ISTI-CNR, Pisa, ItalyVlastislav Dohnal, Masaryk University, Brno, Czech Republic10:30 – 11:00 L OBBYC OFFEE B REAK11:00 – 12:30 R OOMS 310 A AND BAM T UTORIALS IIT1: Introduction to Security-enhanced Linux(SELinux)Dr. Haklin Kimm, Professor, omputer Science Department, ast Stroudsburg University of Pennsylvania, USAT2: Similarity Search - The Metric Space Approach Pavel Zezula, Masaryk University, Brno, Czech RepublicGiuseppe Amato, ISTI-CNR, Pisa, ItalyVlastislav Dohnal, Masaryk University, Brno, Czech Republic 12:00 – 13:30 H OSU (L AKE) F OOD-MALL,1ST F LOORL UNCH B REAK13:30 – 15:00 R OOM 310 APM T UTORIAL IT3: Introduction to OWL Ontology Developmentand OWL ReasoningYoung-Tack Park, Professor, School of Computing, SoongsilUniversity,Seoul, Korea15:00 – 15:30 L OBBYC OFFEE B REAK15:30 – 17:00 R OOM 310 APM T UTORIAL IIT3: Introduction to OWL Ontology Developmentand OWL ReasoningYoung-Tack Park, Professor, School of Computing, SoongsilUniversity,Seoul, Korea18:00 – 19:00 R OOM 311A SAC 2007 R EVIEW M EETINGM ONDAY M ARCH 12, 200708:00 – 17:00 L OBBYR EGISTRATION08:30 – 09:00 R OOM 310O PENING R EMARKS09:00 – 10:00 R OOM 310K EYNOTE A DDRESSA New DBMS Architecture for DB-IRIntegrationDr. Whang, Kyu-YoungDirector of Advanced Information TechnologyResearch CenterKorea Advanced Institute of Science andTechnologyDaejeon, Korea10:00 – 10:30 L OBBYC OFFEE B REAK10:30 – 12:00 R OOM 310A(DS) Data StreamsJoao Gama, University of Porto (UP), Portugal RFID Data Management for Effective ObjectsTrackingElioMasciari, CNR, ItalyA Priority Random Sampling Algorithm for Time-based Sliding Windows over Weighted StreamingDataZhang Longbo, Northwestern Polytechnical University, China Li Zhanhuai, Northwestern Polytechnical University, ChinaZhao Yiqiang, Shandong University of Technology, ChinaMin Yu, Northwestern Polytechnical University, China Zhang Yang, Northwest A&F University, ChinaOLINDDA: A Cluster-based Approach forDetecting Novelty and Concept Drift in DataStreamsEduardo Spinosa, University of Sao Paulo (USP), BrazilAndré Carvalho, University of Sao Paulo (USP), Brazil Joao Gama, University of Porto (UP), PortugalA Self-Organizing Neural Network for DetectingNoveltiesMarcelo Albertini, Universidade de Sao Paulo, BrazilRodrigo Mello, Universidade de São Paulo, Brazil10:30 – 12:00 R OOM 310B (AI) Artificial Intelligence, ComputationalLogic and Image AnalysisChih-Cheng Hung, Southern Polytechnic State University, USA Toward a First-Order Extension of Prolog'sUnification using CHRKhalil Djelloul, University of Ulm, GermanyThi-Bich-Hanh Dao, University d'Orléans, FranceThom Fruehwirth, University of Ulm, GermanyA Framework for Prioritized Reasoning Based onthe Choice EvaluationLuciano Caroprese, University of Calabria, ItalyIrina Trubitsyna, University of Calabria, ItalyEster Zumpano, University of Calabria, ItalyA Randomized Knot Insertion Algorithm for Outline Capture of Planar Images using CubicSplineMuhammad Sarfraz, King Fahd University of Petroleum andMinerals, Saudi ArabiaAiman Rashid, King Fahd University of Petroleum and Minerals,Saudi ArabiaEstraction of Arabic Words from Complex ColorImagesRadwa Fathalla, AAST, EgyptYasser El Sonbaty, AAST College of Computing, Egypt Mohamed Ismail, Alexandria University, Egypt10:30 – 12:00 R OOM 310C (PL) Programming LanguagesMarjan Mernik, University of Maribor, Slovenia Implementing Type-Based Constructive Negation Lunjin Lu, Oakland University, USATowards Resource-Certified Software: A Formal Cost Model for Time and its Application to anImage-Processing ExampleArmelle Bonenfant, University of St Andrews, UKZehzi Chen, Heriot-Watt University, UKKevin Hammond, Univestiy of St Andrews, UKGreg Michaelson, Heriot-Watt University, UKAndy Wallace, Heriot-Watt University, UKIain Wallace, Heriot-Watt University, UK。
DLS 样品制备指南
Guide for DLS sample preparationEric Farrell & Jean-Luc Brousseau Ph.D.NomenclatureThe term solvent refers the pure solvent used toprepare the diluent. Examples of solvents are toluene or water. The diluent may also be referred to as the liquid in DLS textbooks. Diluents are solvent with additives, for example a 10% by weight methanol in water or a 10 mM KNO 3 salt in DI water solution. The samples to be analyzed by DLS will be prepared in the liquid. The solution or suspension to be measured is the sample in the liquid. Although there are differences, the terms suspended and dissolved will be used interchangeably in this document.Aqueous measurementsParticle size measurements by DLS should not beconducted in pure de-ionized (DI) water, as the electrical double layer surrounding the particles will have long distance interaction. The size measured in DI water will usually be too big by 2 to 10 nm due to the electrostatic interaction between the particles. To screen any charge on the particles, it is a good idea to measure in water with a trace amount of salt. The ions with opposite charge will condense around the particle, screening long distance electrostatic interactions. A general salt like NaCl can be used but often the chloride ions are too aggressive and may react with the particles or adsorb to their surface. We recommend the use of KNO 3 for aqueous diluents. A concentration of 10 mM KNO 3 is ideal for all concentrations of particles.Diluent / LiquidIf the liquid is a pure solvent like toluene, thepurest solvent possible shouldbe sourced. Non-polar solvents do not usually dissolve or carry dust. These canThis document is intended to help the user determine the best way to prepare samples for dynamic light scattering (DLS) measurements. If the user wants to know how to prepare the 92nm latex standard for best measurements, please refer to the document entitled "A Guide toProper Sample Preparation: Electrostatically-Stabilized Nanoparticles in Water.”be used as is, as any filtering and manipulation will only risk adding dust. If the solvent is polar, then chances are that filtering will help remove dust in the solvent. In the case of aqueous diluents like KNO 3 in water, filtering is needed as salts are notoriously full of dust. It is always a good practice to filter the aqueous diluent using a 0.1 or 0.2 micrometer filter that has been previously rinsed according to the manufacturer’s practice. Rinsing the filter is an important step that should not be omitted.Dry sample dissolutionIf your sample is a dry powder, it will need to bedissolved or suspended before it can be measured. If the sample is a protein, the solution should not be stirred too aggressively and should never be sonicated. If the sample is sturdy, then sonication, vortex, spinning and other methods can be used to dissolve/suspend the sample. It is for the user to figure out the dissolution time and method. Experiment with different dissolution methods and time. In the case of a sample that disperses quickly, the time to disperse with agitation may be only a few minutes. Large polymers may require more than 24 hours to completely dissolve into solution. When preparing your sample, take a good look at how the sample is dispersing. For further information on preparing stable suspensions from dry powders, refer to ISO standard 14887 entitled “Sample preparation – Dispersing procedures for powders in liquids.”Liquid sample preparationFor liquid samples, the sample may need to bediluted. It is ideal to dilute the sample in the same exact liquid it was originally prepared in, using the same concentrations of additives (i.e. salts, surfactants,dispersing agents) if any were present. In highly concentrated liquid samples, the samples may appear opaque or milky-white. If the sample is highly concentrated, the sample should be diluted in the liquid of choice. Usually putting a drop of the neat sample in 20 mL of liquid or doing a 1:1000 dilution should be sufficient.ConcentrationsSolutions prepared for DLS will need to be clearto very slightly hazy. Although the instrument can measure solutions at concentration up to 40% and possibly more, the size measured in these cases will be wrong. It is not a DLS measurement if the sample is not clear. The Stokes-Einstein equation applies to infinitely dilute solutions. Highly concentrated measurements can be made for diffusing wave spectroscopy (DWS) and other applications. For DLS particle sizing, the sample needs to be water clear to very slightly hazy. If the solution is white or too hazy, it should be diluted further before attempting a DLS size measurement. If the sample is too concentrated, the measured size of your particles will be inaccurate due to multiple scattering or viscosity effects. Multiple scattering occurs when the particle concentration is high enough where the light scattered from a single particle is re-scattered by others in the suspension. This will cause your measured particle size to be artificially low. Viscosity effects occur when the volume of sample added to the liquid is enough such that it will alter the viscosity of the liquid. If the liquid viscosity is wrong due to viscosity effects, your measured particle size will also be wrong. A simple schematic is shown to describe the effects ofmeasuring at high concentration. Depending on the sample and the liquid, a size reduction or a size increase can be seen when theconcentration is too high for DLS. At extremely high concentrations, this phenomenom usually reverses.When the solution is ready for analysis, it should be inspected for particles at the bottom of the cuvette. If there are particles at the bottom of the cuvette, the sample distribution will not be accurate, as the large settled particles will not be measured, resulting in an inaccurate size distribution. Samples with large particles settling are either not dispersed correctly (wrong pH, not enough sonication, not enough dilution time, etc.) or are not suitable for DLS. For low density particles, samples may exhibit the opposite behavior. Creaming is the term used for large particles of lower density than the solvent reaching the top of the sample. If the sample is creaming, there is either a dispersion problem or the particles are too big to be measured by DLS.When the solution is ready for analysis and placed in the cuvette, care should be taken to avoid bubbles that may form on the walls of the cuvette. Slowly tilting or tapping the cuvette on a hard surface may help also.Colored solutionsIn the case of colored solutions, as long as the laser light is not absorbed completely by the solution, the sample can be measured. Colored samples and fluorescing samples may be harder to measure. Often the sample is available without the fluorescent dye or the absorbing chromophore. In this case it will be easier to measure the sample without the dye/chromophore present. To distinguish the scattering from the inherent color of a sample, try reading text through the sample. If the text can be read, then the scattering is not creating most of the opacity.Filtering the solutionIf the solution is to be filtered, keep in mind that the size distribution may be changed if the particles are removed by the filter. A good rule of thumb is to use a pore size (filter size) 3 times larger than the largest size to be measured. For DLS a 5 micrometer filter can be used in most cases. Always verify that the largest size measured is smaller than the filter pore size by a factor of 3. Always rinse the filter prior to use. These recommendations are valid for all filters except the Whatman Anotop series filters. For Whatman Anotop series filters, pass the first drop of sample to waste.MeasurementsOnce the solution is homogenous and ready for DLS measurement, the solution can be placed in the instrument. For the novice, there are two ways of checking that the concentration is not too high and that the DLS measurement will be valid.1) Count rate check: For an instrument equipped with an APD, the count rate for the scattering intensity of the dilute solution should be less than 2 Mcps (2,000,000 count per seconds) with the intensity maximized. Note that measurements SHOULD NOT be made at this high count rate. The maximum count rate for the measurement should be 500-600 kcps. The attenuator will need to be adjusted after measuring the scattering intensity on maximum intensity in most cases, especially if the count rate is greater than 600 kcps.2) Dilution check: The second way to verify that the concentration is suitable for DLS measurements is to dilute your sample by 50% after the first measurement. If the size of the dilution is the same as the size of the more concentrated measurement and the count rate is reduced by a factor of 2, then the first measurement concentration was low enough.After many measurements it will be easy to assess the concentration of the solution to be measured.Forbest practices for measuring samples size, please refer to the document “Guide for DLS measurements”.Nanoparticle, Protein, & Polymer Characterization。
2005-A global Malmquist productivity index
A global Malmquist productivity indexJesu ´s T.Pastor a ,C.A.Knox Lovell b ,TaCentro de Investigacio ´n Operativa,Universidad Miguel Herna ´ndez,03206Elche (Alicante),SpainbDepartment of Economics,University of Georgia,Athens,GA 30602,USA Received 2June 2004;received in revised form 24January 2005;accepted 16February 2005Available online 23May 2005AbstractThe geometric mean Malmquist productivity index is not circular,and its adjacent period components can provide different measures of productivity change.We propose a global Malmquist productivity index that is circular,and that gives a single measure of productivity change.D 2005Elsevier B.V .All rights reserved.Keywords:Malmquist productivity index;Circularity JEL classification:C43;D24;O471.IntroductionThe geometric mean form of the contemporaneous Malmquist productivity index,introduced by Caves et al.(1982),is not circular.Whether this is a serious problem depends on the powers of persuasion of Fisher (1922),who dismissed the test,and Frisch (1936),who endorsed it.The index averages two possibly disparate measures of productivity change.Fa ¨re and Grosskopf (1996)state sufficient conditions on the adjacent period technologies for the index to satisfy circularity,and to average the same measures of productivity change.When linear programming techniques are used to compute and decompose the index,infeasibility can occur.Whether this is a serious problem depends on0165-1765/$-see front matter D 2005Elsevier B.V .All rights reserved.doi:10.1016/j.econlet.2005.02.013T Corresponding author.Tel.:+17065423689;fax:+17065423376.E-mail address:knox@ (C.A.K.Lovell).Economics Letters 88(2005)266–271/locate/econbasethe structure of the data.Xue and Harker(2002)provide necessary and sufficient conditions on the datafor LP infeasibility not to occur.We demonstrate that the source of all three problems is the specification of adjacent periodtechnologies in the construction of the index.We show that it is possible to specify a base periodtechnology in a way that solves all three problems,without having to impose restrictive conditions oneither the technologies or the data.Berg et al.(1992)proposed an index that compares adjacent period data using technology from a baseperiod.This index satisfies circularity and generates a single measure of productivity change,but it paysfor circularity with base period dependence,and it remains susceptible to LP infeasibility.Shestalova(2003)proposed an index having as its base a sequential technology formed from data ofall producers in all periods up to and including the two periods being compared.This index is immune toLP infeasibility,and it generates a single measure of productivity change,but it fails circularity and itprecludes technical regress.Thus no currently available Malmquist productivity index solves all three problems.We propose anew global index with technology formed from data of all producers in all periods.This index satisfiescircularity,it generates a single measure of productivity change,it allows technical regress,and it isimmune to LP infeasibility.In Section2we introduce and decompose the circular global index.Its efficiency change componentis the same as that of the contemporaneous index,but its technical change component is new.In Section3we relate it to the contemporaneous index.In Section4we provide an empirical illustration.Section5concludes.2.The global Malmquist productivity indexConsider a panel of i=1,...,I producers and t=1,...,T time periods.Producers use inputs x a R N+toproduce outputs y a R P+.We define two technologies.A contemporaneous benchmark technology isdefined as T c t={(x t,y t)|x t can produce y t}with k T c t=T c t,t=1,...,T,k N0.A global benchmarktechnology is defined as T c G=conv{T c1v...v T c T}.The subscript b c Q indicates that both benchmark technologies satisfy constant returns to scale.A contemporaneous Malmquist productivity index is defined on T c s asM scx t;y t;x tþ1;y tþ1ÀÁ¼D scx tþ1;y tþ1ðÞD scx t;y tðÞ;ð1Þwhere the output distance functions D c s(x,y)=min{/N0|(x,y//)a T c s},s=t,t+1.Since M c t(x t,y t,x t+1, y t+1)p M c t+1(x t,y t,x t+1,y t+1)without restrictions on the two technologies,the contemporaneous index is typically defined in geometric mean form as M c(x t,y t,x t+1,y t+1)=[M c t(x t,y t,x t+1,y t+1)ÂM c t+1(x t,y t,x t+1, y t+1)]1/2.A global Malmquist productivity index is defined on T c G asM Gcx t;y t;x tþ1;y tþ1ÀÁ¼D Gcx tþ1;y tþ1ðÞD Gcx t;y tðÞ;ð2Þwhere the output distance functions D c G(x,y)=min{/N0|(x,y//)a T c G}.J.T.Pastor,C.A.K.Lovell/Economics Letters88(2005)266–271267Both indexes compare (x t +1,y t +1)to (x t ,y t ),but they use different benchmarks.Since there is only one global benchmark technology,there is no need to resort to the geometric mean convention when defining the global index.M cGdecomposes as M G c x t ;y t ;x t þ1;f y t þ1ÀÁ¼D t þ1c x t þ1;y t þ1ðÞD t c x t ;y t ðÞÂD G c x t þ1;y t þ1ðÞD t þ1c x t þ1;y t þ1ðÞÂD t cx t ;y t ðÞD Gc x t ;y t ðÞ&'¼TE t þ1c x t þ1;y t þ1ðÞTE t c x t ;y t ðÞÂD G c Àx t þ1;y t þ1=D t þ1c x t þ1;y t þ1ðÞÁD G c x t ;y t =D t cx t ;y t ðÞÀÁ()¼EC c ÂBPG G ;t þ1cx t þ1;y t þ1ðÞBPG cx t ;y tðÞ()¼EC c ÂBPC c ;ð3Þwhere EC c is the usual efficiency change indicator and BPG c G,s V 1is a best practice gap between T c Gand T c s measured along rays (x s ,y s),s =t ,t +1.BPC c is the change in BPG c ,and provides a new measure of technical change.BPC c f 1indicates whether the benchmark technology in period t +1in the region[(x t +1,y t +1/D ct +1(x t +1,y t +1))]is closer to or farther away from the global benchmark technology than is the benchmark technology in period t in the region [(x t ,y t /D ct (x t ,y t ))].M c G has four virtues.First,like any fixed base index,M cGis circular,and since EC c is circular,so is BPC c .Second,each provides a single measure,with no need to take the geometric mean of disparate adjacent period measures.Third,but not shown here,the decomposition in (3)can be extended to generate a three-way decomposition that is structurally identical to the Ray and Desli (1997)decomposition of the contemporaneous index.M cGand M c share a common efficiency change component,but they have different technical change and scale components,and so M c Gp M c without restrictions on the technologies.Finally,the technical change and scale components of M c Gare immune to the LP infeasibility problem that plagues these components of M c .paring the global and contemporaneous indexes The ratioM G c =M c¼M G c =M t þ1cÀÁÂM G c =M t cÀÁÂÃ1=2¼D G cx t þ1;y t þ1=D t þ1c x t þ1;y t þ1ðÞÀÁD G c x t ;y t =D t þ1c x t ;y t ðÞÀÁ"#ÂD G c x t þ1;y t þ1=D t c x t þ1;y t þ1ðÞÀÁD G c x t ;y t =D t c x t ;y t ðÞÀÁ"#()1=2¼BPG G ;t þ1cx t þ1;y t þ1ðÞBPG G ;t þ1cx t ;y tðÞ"#ÂBPG G ;t c xt þ1;y t þ1ðÞBPG G ;t c x t ;y tðÞ"#()1=2ð4Þis the geometric mean of two terms,each being a ratio of benchmark technology gaps along differentrays.M c G /M c f 1as projections onto T c t and T c t +1of period t +1data are closer to,equidistant from,orfarther away from T c G than projections onto T c t and T ct +1of period t data are.J.T.Pastor,C.A.K.Lovell /Economics Letters 88(2005)266–271268J.T.Pastor,C.A.K.Lovell/Economics Letters88(2005)266–271269 Table1Electricity generation data,annual means1977198219871992 Output(000MW h)13,70013,86016,18017,270 Labor(#FTE)1373179719952021 Fuel(billion BTU)1288144116671824 Capital(To¨rnqvist)44,756211,622371,041396,386 M c G=M c if BPG c G,s(x t+1,y t+1)=BPG c G,s(x t,y t),s=t,t+1.From the first equality in(4),this condition is equivalent to the condition M c G=M c s,s=t,t+1.If this condition holds for all s,it is equivalent to the condition M c t=M c1for all t.Althin(2001)has shown that a sufficient condition for base period independence is that technical change be Hicks output-neutral(HON).Hence HON is also sufficient for M c G=M c.4.An empirical illustrationWe summarize an application intended to illustrate the behavior of M c G,and to compare its performance with that of M c.We analyze a panel of93US electricity generating firms in four years (1977,1982,1997,1992).The firms use labor(FTE employees),fuel(BTUs of energy)and capital(a multilateral To¨rnqvist index)to generate electricity(net generation in MW h).The data are summarized in Table1.Electricity generation increased by proportionately less than each input did.The main cause of the rapid increase in the capital input was the enactment of environmental regulations mandating the installation of pollution abatement equipment.We are unable to disaggregate the capital input into its productive and abatement components.Empirical findings are summarized in Table2.The first three rows report decomposition(3)of M c G, and the final three rows report M c and its two adjacent period components.Columns correspond to time periods.M c G shows a large productivity decline from1977to1982,followed by weak productivity growth. Cumulative productivity in1992was25%lower than in1977.M c G calculated using1992and1977data generates the same value,verifying that it is circular.The efficiency change component EC c of M c G(and M c)is also circular,and cumulates to an18% improvement.Best practice change,BPC c,is also circular,and declined by35%.Capital investment in Table2Global and contemporaneous Malmquist productivity indexes1977–19821982–19871987–1992Cumulative productivity1977–1992 M c G0.685 1.064 1.0390.7570.757EC c 1.163 1.0890.929 1.176 1.176 BPC c0.5890.977 1.1180.6440.644M c0.4310.895 1.0390.4000.592M c t0.7130.902 1.0530.678 1.333M c t+10.2600.887 1.0240.2360.263pollution abatement equipment generated cleaner air but not more electricity.Consequently catching up with deteriorating best practice was relatively easy.Turning to the contemporaneous index M c reported in the final three rows,the story is not so clear.Cumulative productivity in 1992was 60%lower than in 1977.However calculating M c using 1992and 1977data generates a smaller 40%decline,verifying that M c is not circular.Neither figure is close to the25%decline reported by M cG,verifying that technical change was not HON,but (pollution abatement)capital-using.The lack of circularity is reflected in the frequently large differences between M ct and M c t +1,which give conflicting signals when computed using 1992and 1977data,with M c tsignaling productivitygrowth and M ct +1signaling productivity decline.Although not reported in Table 2,we have calculated three-way decompositions of M cG and M c .All three components of M c G are circular,and LP infeasibility does not occur.In contrast,the technical change and scale components of M c are not circular,and infeasibility occurs for 13observations.The circular global index M cGtells a single story about productivity change,and its decomposition is intuitively appealing in light of what we know about the industry during the cking circularity,M c and its two adjacent period components tell different stories that are often contradictory.Thedifferences between M cGand M c are a consequence of the capital-using bias of technical change,which was regressive due to the mandated installation of pollution abatement equipment,augmented perhaps by the rate base padding that was prevalent during the period.5.ConclusionsThe contemporaneous Malmquist productivity index is not circular,its adjacent period components can give conflicting signals,and it is susceptible to LP infeasibility.The global Malmquist productivity index and each of its components is circular,it provides single measures of productivity change and its components,and it is immune to LP infeasibility.The global index decomposes into the same sources of productivity change as the contemporaneous index does.A sufficient condition for equality of the two indexes,and their respective components,is Hicks output neutrality of technical change.The global index must be recomputed when a new time period is incorporated.Diewert’s (1987)assertion that b ...economic history has to be rewritten ...Q when new data are incorporated is the base period dependency problem revisited.The problem can be serious when using base periods t =1and t =T ,but it is likely to be benign when using global base periods {1,...,T }and {1,...,T +1}.While new data may change the global frontier,the rewriting of history is likely to be quantitative rather than qualitative.ReferencesAlthin,R.,2001.Measurement of productivity changes:two Malmquist index approaches.Journal of Productivity Analysis 16,107–128.Berg,S.A.,Førsund,F.R.,Jansen,E.S.,1992.Malmquist indices of productivity growth during the deregulation of Norwegian banking,1980–89.Scandinavian Journal of Economics 94,211–228(Supplement).Caves,D.W.,Christensen,L.R.,Diewert,W.E.,1982.The economic theory of index numbers and the measurement of input output,and productivity.Econometrica 50,1393–1414.J.T.Pastor,C.A.K.Lovell /Economics Letters 88(2005)266–271270J.T.Pastor,C.A.K.Lovell/Economics Letters88(2005)266–271271 Diewert,W.E.,1987.Index numbers.In:Eatwell,J.,Milgate,M.,Newman,P.(Eds.),The New Palgrave:A Dictionary of Economics,vol.2.The Macmillan Press,New York.Fa¨re,R.,Grosskopf,S.,1996.Intertemporal Production Frontiers:With Dynamic DEA.Kluwer Academic Publishers,Boston. Fisher,I.,1922.The Making of Index Numbers.Houghton Mifflin,Boston.Frisch,R.,1936.Annual survey of general economic theory:the problem of index numbers.Econometrica4,1–38.Ray,S.C.,Desli,E.,1997.Productivity growth,technical progress,and efficiency change in industrialized countries:comment.American Economic Review87,1033–1039.Shestalova,V.,2003.Sequential Malmquist indices of productivity growth:an application to OECD industrial activities.Journal of Productivity Analysis19,211–226.Xue,M.,Harker,P.T.,2002.Note:ranking DMUs with infeasible super-efficiency in DEA models.Management Science48, 705–710.。
Place Marketing as Governance Strategy An Assessment of Obstacles in Place Marketing
Jasper Eshuis is assistant professor inthe Department of Public Administration at Erasmus University Rotterdam. His research focuses on the governance of complex systems, with a special interest in public branding and place marketing. Together with Erik-Hans Klijn, he recently published Branding in Governance and Public Management (Routledge, 2012).E-mail: eshuis@fsw.eur.nlErik Braun is senior researcher and lecturer in urban economics and city mar-keting at the Erasmus School of Economics, Erasmus University Rotterdam. He published his doctoral dissertation on city marketing in 2008. His research interests include the application of marketing and branding concepts by cities and regions, place brand management, place brand perceptions, and the governance of place marketing as well as a broad range of urban economic issues.E-mail: braun@ese.eur.nlErik-Hans Klijn is professor in the Department of Public Administration at Erasmus University Rotterdam. His research activities focus on complex decision making, network management, public–private partnerships, branding, and the impact of media on complex decision making. He has published extensively in international journals and is author of Managing Uncertainties in Networks (with Joop F . M. Koppenjan, Routledge, 2004) and Branding in Governance and Public Management (with Jasper Eshuis, Routledge, 2012).E-mail: klijn@fsw.eur.nlPlace Marketing as Governance Strategy: An Assessment of Obstacles in Place Marketing and Their Effects on Attracting Target Groups 507Public Administration Review , Vol. 73, Iss. 3, pp. 507–516. © 2013 by The American Society for Public Administration. DOI: 10.1111/puar.12044.Jasper EshuisErik Braun Erik-Hans KlijnErasmus University Rotterdam, The NetherlandsPlace marketing is increasingly being used as a g overnance strategy for managing perceptions about regions, cit-ies, and towns. What are the most important obstacles to implementing place marketing? Based on a survey of 274 public managers involved in place marketing in the Netherlands, this article analyzes the main obstacles as perceived by public managers. It also analyzes the eff ects of obstacles on perceived results of place marketing in terms of attracting target groups. A factor analysis of a variety of obstacles investigated in the survey shows three clearly demarcated obstacles: administrative obstacles within municipalities, obstacles in developing the substance of marketing campaigns, and political obstacles. Obstacles in developing the substance of the marketing campaigns have signifi cant eff ects on the results of place marketing in terms of attracting stakeholders, whereas the two other obstacles have no signifi cant infl uence.Place marketing has become a strategy widely deployed by munici-palities and regional authorities in the governance of cities, towns, and regions. It is used to increase the competitive-ness of places and attract target groups such as tourists, new residents, and investors (see, e.g., Bennett and Savani 2003; Braun 2008; Hospers 2006). It may include promotion and creating a positive image, as well as product development in the sense of developing the place in a way thatresponds to the demands of target groups (Greenberg 2008; Kavaratzis 2004; Kotler and Gertner 2002). Although place marketing has been common in the United States for more than three decades (Gold and Ward 1994), it is a relatively new and growing phe-nomenon in Europe and other parts of the world. Municipalities in smaller cities, such as Hasselt in Belgium or Randers in Denmark, have followed the lead of major cities such as London (“We are Londoners”) or Montevideo (“Montevideo for All”) (see Dinnie 2011).Place Marketing as Upcoming GovernanceStrategy Th e upsurge of place marketing can be understood in the context of the wider governance trend of introduc-ing commercial practices and private sector manage-ment styles (cf. Pierre and Peters 2000). Market-based reforms have taken place under the heading of New Public Management (see, e.g., Barzelay 2001; Pollitt, Van Th iel, and Homburg 2007), which stresses the importance of increasing effi ciency in the public sector by introducing competition and approaching citizens more like customers. Customer consciousness and customer care training have become common in pub-lic services, as well as strategic marketing approaches, including business plans, market segmentation, and branding (Eshuis and Klijn 2012; Walsh 1994). In line with these governance trends, municipalities have introduced more marketing-led urban governancestrategies (Greenberg 2008). Th is has led not only to market-ing plans but also to the restruc-turing of spatial, economic, and fi scal policies. Greenberg (2008) describes how the branding of New York was entwined withpro-business restructuring of(fi scal) policies. Place market-ing has evolved from applyingparticular promotional techniques for purposes such as increasing tourism to marketing as an integral part of urban governance.However, the application of place marketing has not always been smooth or successful. Place marketers face multiple constraints or obstacles when trying to apply marketing strategies in a public sector context. Like other forms of governance, place marketing involvesmany diff erent actors who may disagree about, for example, marketing instruments or the brand that best captures the aspired identity of the city. Another issue in place marketing is that a place is a complex “product” that may be diffi cult to market. Although the literature mentions quite a few diffi culties in placePlace Marketing as Governance Strategy: An Assessmentof Obstacles in Place Marketing and Th eir Eff ectson Attracting Target GroupsPlace marketing has evolved from applying particular pro-motional techniques for pur-poses such as increasing tourism to marketing as an integral part of urban governance.marketing, no comprehensive research has been undertaken on the obstacles that public managers encounter in place marketing. Obstacles in Place Marketing and Their Effects on OutcomesTh is article aims to address this knowledge gap. Th e goal is to empirically determine the most important obstacles in place market-ing, according to professionals and politicians involved in place marketing, and their eff ects on the outcome of place marketing in terms of attracting target groups.T wo main research questions are addressed:1. What are the main obstacles in place marketing?2. What is the relationship between obstacles and outcomes ofplace marketing in terms of attracting target groups?Th e article is structured as follows: Th e next section lays a theoreti-cal basis by defi ning what place marketing is. Th e literature on risks, limits, and obstacles to place marketing is then discussed. Th e next section describes the research design and methodology. Th is is fol-lowed by the empirical fi ndings. Th e article ends with a discussion and conclusions.Place Marketing: What Is It?Place marketing involves the application of marketing instruments to geographic locations, such as nations, cities, regions, and com-munities. Th e place marketing literature mentions a multiplicity of activities, instruments, and strategies under the heading of the mar-keting mix that can be applied to places. Th erefore, most scholars consider that marketing is broader than promotion only (see Braun 2008; Eshuis and Edelenbos 2008; Hospers 2006). Ashworth and Voogd (1990) develop a geographic marketing mix that includes not only promotional measures, but also spatial-functional measures, organizational measures,and fi nancial measures intended to improvethe place and its management. Kotler, Haider,and Rein (1993) include activities aimedat improving design and service delivery, aswell as developing attractions. Th e emphasison multiple activities and strategies in placemarketing refl ects the idea that mere promo-tion, without developing the product and themanagement, is not very useful if one wantsto attract people or organizations to a placeand increase competitiveness.Place Marketing: More than Communicating Favorable Images Place marketing is more than just developing favorable images and communicating them to the diff erent target groups; it is not only about “selling” an image. Kotler emphasizes time and again that marketing is about fulfi lling consumer needs (e.g., Kotler et al. 1999). Th us, place marketing is about developing a place that fi ts the needs and wants of citizens, visitors, and investors. Marketing is about responsiveness more than persuasion, although persuasion is an important part of place marketing. Th e idea behind the broader marketing approach is that marketing is much more eff ective if itis targeted at what stakeholders want. Here, marketing is aboutnot only sending messages but also receiving messages. Marketing then becomes a matter of developing the place that people want and applying elements of policy making, urban planning, and place development (or product development in marketing terms). Th is makes place marketing a special governance strategy that explicitly includes the management of wider processes of urban development. Th is view is shared by scholars such as Van den Berg and Braun, who state that “urban place marketing can be seen as a managerial principle in which thinking in terms of customers and the market is central as well as a toolbox with applicable insights and techniques” (1999, 993). Here, place marketing is a way of thinking and doing that emphasizes a consumer orientation or, to put it slightly diff er-ently, a demand-driven orientation. Th is article follows this line.In the words of Braun, place or city marketing is defi ned as “the coordinated use of marketing tools supported by a shared customer-oriented philosophy, for creating, communicating, delivering, and exchanging urban off erings that have value for the city’s customers and the city’s community at large” (2008, 43). However, place mar-keting as a governance strategy also faces constraints and obstacles, summarized as obstacles in the next section.Exploring Obstacles in Place MarketingObstacles in place marketing arise from the governance environ-ment in which it is employed and from obstacles in place marketing strategies themselves. Th e literature on place marketing and brand-ing mentions a long list of obstacles to undertaking place marketing and achieving good outcomes from it. Th e governance literature provides insight into obstacles relating to the working of public organizations and political processes in the public sector. In this sec-tion, the main obstacles to place marketing are briefl y discussed.Th e fi rst thing stressed in the literature on both governance (Pierre 2000) and place marketing (Klijn, Eshuis, and Braun 2012) is that,in a governance context, multiple public andprivate parties are involved: for example, thetourist board, hotels, museums, some majorcompanies, and the municipality. A character-istic of governance processes is that actors mayhave diff erent or even confl icting preferences(Koppenjan and Klijn 2004; McGuire andAgranoff 2011; Pierre 2000). Diff erent actorsmay have diff erent perceptions about theaims, strategies, and target groups of placemarketing campaigns. Th ere may be diff er-ences between public and private actors, butvarious public departments also often havediff erent interests and favor diff erent solutions. Government institu-tions are themselves fragmented, as much of the literature on gov-ernance emphasizes (e.g., Atkinson and Coleman 1992). Th is may create obstacles. McGuire and Agranoff (2011) state that depart-mental separation is a barrier to becoming a conductive organization that is capable of calibrating to the needs of its customers and the marketplace. For example, decision making can be hampered if the department of housing favors a campaign that positions the city as a nice and peaceful residential area, whereas the economics depart-ment wants to stress industrial investment opportunities.Another major barrier to interorganizational collaboration in govern-ance is what Bardach (1996) calls “turf problems,” where turf refersTh e emphasis on multipleactivities and strategies in placemarketing refl ects the ideathat mere promotion, withoutdeveloping the product and themanagement, is not very usefulif one wants to attract peopleor organizations to a place andincrease competitiveness.508Public Administration Review • May | June 2013to the domain over which an agency is responsible and exercises legitimate authority. When municipal departments such as planning departments are confronted with new place marketing agencies that want to make city development more market oriented, they may be inclined to protect their turf and resist collaboration. Th is problem is more pressing when a governance strategy, as is the case with place marketing, requires coordination and anintegrated approach to be (more) successful(Braun 2008). Place branding is a relativelynew policy fi eld and thus probably suff ersfrom fragmentation and a lack of coordinationwith other policy activities. Th is fragmentationmay hinder eff ective place marketing strate-gies and implementation (Kavaratzis 2009),as is also known from public administrationimplementation studies (Pressman and Wildavsky 1973). Th us, it is expected that one of the obstacles will relate to this fragmentation and lack of coordination within government institutions. Governance is not only about administrators and administrative bod-ies, but as much about politicians and the public. Although scholars have posited shifts of power from representative political bodies to governance networks (Hanf and Scharpf 1978; Koppenjan and Klijn 2004; Pierre 2000), political bodies still exercise a degree of control. In governance, politicians are not political executives at the apex of a government hierarchy, but they do play a role in steering policy development, determining budgets, and overseeing implementation (Scharpf 1997). Policy development in governance, including place marketing policies, requires political support and suffi cient approval by the citizenry. A lack of support among political representatives is an important barrier to good results in governance networks (Klijn and Koppenjan 2000; McGuire and Agranoff 2011).Th e issue of lack of political support has been recognized as a prob-lem in the literature on place marketing. Braun (2008) fi nds, in a study comparing four European cities, that a lack of political prior-ity for place marketing can hinder its proper embedding in wider urban governance. Lack of political priority may also lead to a lack of fi nancial resources for place marketing.Apart from obstacles stemming from the governance context of place marketing, there are also constraints rooted in the nature of place marketing itself. Th e literature on place marketing highlights diffi culties in relation to the particular nature of places and the users of places. As Ward and Gold argue, “it is not readily apparent what the product actually is, nor how the consumption of place occurs. Th ough marketing practices make places into commodities, they are in reality complex packages of goods, services and experiences that are consumed in many diff erent ways” (1994, 9). For example, a city may be seen as a tourist destination, but at the same time, it is a place of residence for many people or a location through which they need to travel when going elsewhere. In short, a place is a complex product with multiple identities (Kavaratzis and Ashworth 2005). Th e identities of a city involve, for example, the city’s history and its historic center, but also its information and communication technologies and gaming industry. Th us, the city can have diff er-ent identities for diff erent target groups. T ourists, for example, may value the historic center, whereas private companies may value the information and communication technology industry. A city’s multiple coexisting identities make it diffi cult to develop appropri-ate place marketing campaigns.Other obstacles arise from the multiplicity of uses of places and the public character of places. Local governments and other stakehold-ers may use brands to appeal to diff erent groups and evoke diff erentassociations with them, but this is not easy.A brand that is suitable for one group (e.g.,tourists) may not suit other groups (e.g.,residents) (Bennett and Savani 2003). In acommercial setting, marketers can choosetarget groups and ignore others, but in a pub-lic context, it may be illegitimate or impos-sible to ignore groups of residents, voters,or businesses (see also Eshuis and Edwards 2012). It often proves diffi cult to create marketing plans that fi t the preferences of all stakeholders. Stigel and Frimann (2006) fi nd this limitation of place marketing in their study of the branding of two Danish towns. Th ey conclude that the wish to arrive at a consensus about brand identities can easily lead to brands with only very gen-eral and nondescript values. Th is inhibits the eff ectiveness of place marketing in terms of creating a distinguishable identity and a clear profi le that makes the place stand out among its competitors.So, in the literature, many obstacles regarding place marketing are discussed. However, almost all of the literature cited earlier is based on case study research, so it remains unclear what the most signifi -cant obstacles are and how diff erent obstacles are interrelated. Th is article addresses this issue through quantitative survey research.Th e obstacles mentioned in the literature have been translated into survey items to identify the obstacles considered most important by practitioners, the underlying factors in the obstacles, and how they relate to outcomes of place marketing.Research Design: Methodology and Methods Organization of the SurveyTh is research is based on data collected for the fi rst National City Marketing Monitor in the Netherlands 2010 (see also Braun et al. 2010; Klijn, Eshuis, and Braun 2012). Th is was a Web-based survey sent to professionals and city administrators involved in the market-ing of cities, towns, and villages. To acquire a reliable set of respond-ents actually involved in Dutch place marketing, the research was carried out in close collaboration with three organizations that provided e-mail addresses of (potential) respondents:• Th e main Dutch network for place marketing in the Neth-erlands, Netwerk City Marketing Nederland. Th is nonprofi tassociation for professionals working in place marketing was es-tablished in 2009. Th e association has no paid employees (except for support occasionally hired for secretarial work and account-ing). Th e fi ve board members are professional marketers whodo this work on a voluntary basis. Th e association facilitates the development of a network of professionals working in place mar-keting by maintaining a Web site and a LinkedIn group (2,300members) and by regularly organizing seminars and conferences on place marketing. Th ose meetings attracted about 1,400 at-tendants over the past three years. For this research, the network provided the e-mail addresses of people who had participated in these events. Th is provided a large set of respondents.Local governments and otherstakeholders may use brands toappeal to diff erent groups andevoke diff erent associations withthem, but this is not easy.Place Marketing as Governance Strategy: An Assessment of Obstacles in Place Marketing and Their Effects on Attracting Target Groups 509• Th e Dutch organization for local and regional tourism offi ces, VVV Nederland, the umbrella organization for municipaltourism offi ces. VVV aims to further tourism and recreationin the Netherlands since it was founded in 1885. Th e national organization has about 40 employees, but more than 1,000employees work in approximately 200 local offi ces (see http://www.nuzakelijk.nl/werk/2298066/medewerkers-vvv-hebben-125-jaar-cao.html). Th e local offi ces engage in destinationmarketing and provide tourists with information about theplace, especially regarding recreational possibilities. Th e VVVs are often largely subsidized by municipalities. For this research, the VVV provided details of respondents within local tourism offi ces who were involved in place marketing.• Th e Dutch association for local governments, VerenigingNederlandse Gemeenten (VNG). Th is is an umbrella organiza-tion for municipalities in the Netherlands that has existed since 1912 and has about 270 employees. It advocates for the inter-ests of municipalities and served as a platform for consultation and knowledge exchange among municipalities. All 408 Dutch municipalities are voluntary members of the VNG. For thisresearch, the VNG provided additional addresses of munici-palities and contact persons. Th is facilitated further researchinto people involved in place marketing.Drawing on the three diff erent lists of respondents and comple-menting this with an additional search for respondents in munici-palities (using an existing network of people working in place marketing) resulted in a reasonably complete data set of 600 people involved in place marketing in the Netherlands. To be sure, it is hard to assess whether the list is complete because of the lack of offi -cial registration of people working in place marketing. However, this extensive search for respondents gives reasonable confi dence that at least a very large proportion of this group is included.During the fi rst round, the survey was sent to the 600 names on the list. Of these, 541 were reached. In all, 274 answered at least partof the survey, giving a response rate of 51 percent. Th e high level of response can be attributed at least partly to the involvement of the aforementioned three organizations and their support for the survey. Survey RespondentsOf the 274 respondents, 168 worked for a municipality, 68 worked for a tourism offi ce, and 38 worked for an organization at arm’s length—usually a foundation involved in place marketing (tourism offi ces often participate in such organizations). Th e respondents had a variety of functions, varying from communication advisor to neighborhood manager, policy advisor, and city alderman. More than 53 percent of the respondents had more than two years’ experi-ence with place marketing.Larger cities are overrepresented in the survey. Th e proportion of municipalities in the Netherlands with fewer than 50,000 inhabit-ants is almost 60 percent, but the number of respondents fromthis group is only 37 percent. Th e largest cities in the Netherlands (more than 250,000 inhabitants) represent only 1.5 percent of all municipalities, whereas almost 13 percent of our respondents come from this group. Th is may not come as a surprise, however, as large municipalities tend to employ more people, including those in place marketing, than small municipalities. Although the sample may not be representative of all municipalities, it can be confi dently claimed to represent the people involved in place marketing thanks to the broad coverage of professionals through their representative organiza-tions and the good response rate (51 percent).Measuring Obstacles in Place MarketingTh rough a combination of inductive and deductive steps, 11 survey items were developed to measure obstacles in place marketing. First, the literature on governance and place marketing was explored to fi nd obstacles rooted in the governance context, as well as obstacles stem-ming from diffi culties in the marketing of places. Th ese obstacles were then organized into categories (interdepartmental issues, impact on product development, political support, budget, reaching and infl uenc-ing citizens, fi t with place identity, and clarity of the brand) and trans-lated into measurable items. Th e list was shown to a place marketer and another professional from the public sector (a policy advisor) to see whether the most important obstacles (as perceived by practitioners and marketers) had been included. Th e items are presented in table 1. Perceptions of the obstacles were measured on a Likert scale (“com-pletely agree,” “agree,” “neither agree nor disagree,” “disagree,” and “completely disagree”).Measuring the Attraction of Target Groups (Dependent Variable)Our dependent variable aims to measure the overall performanceof place marketing in terms of attracting target groups. Overall performance is determined by the degree to which multiple target groups are attracted. Th erefore, we created a composite dependent variable, attracting target groups, which was measured using three items for the three most important target groups in place marketing:1. Place marketing has contributed positively to attractingvisitors.2. Place marketing has contributed positively to attracting newresidents.3. Place marketing has contributed positively to attractingcompanies/fi rms.Each question was measured on a Likert scale ranging from “com-pletely agree” to “completely disagree.”Table 1Items for Measuring ObstaclesItem1The budget for place marketing is too low.2It is diffi cult to reach consensus within the municipality about the place marketing content and strategy.3There is insuffi cient expertise within the municipality.4Policy departments view place marketing as a threat, they do not want place marketing to infl uence their policy.5Various municipal departments do not give marketing much consideration in their communication.6Place marketing has insuffi cient impact on product development.7There is not enough political support.8Place marketing does not really strike a chord with the citizen.9The campaign does not fi t the identity of the municipality.10The campaign does not provide a clear profi le for the municipality.11The intended target groups are not reached suffi ciently.Note: This article uses the term place marketing. The Dutch survey actually used the term city marketing because in the Netherlands, this is the most commonly used term to denote the marketing of places (whether it be cities, towns,v illages, or districts).510Public Administration Review • May | June 2013Factor AnalysisA factor analysis was performed to fi nd the latent variables behind the 11 identifi ed obstacles. In addition, a factor analysis was conducted on the three items listed earlier to derive the dependent variable, attracting target groups. As already mentioned, all 11 items for the obstacles and the three items for the dependent variable were measured on Likert scales. Both factor analyses used the polychoric correlation matrix rather than the standard Pearson correlation matrix to respect the ordinal character of the data, as suggested by Kolenikov and Angeles (2004). Th e literature has for some time been suggesting that it is incorrect to treat ordinal data as interval or ratio variables (Babakus, Ferguson, and Joereskog 1987; Muthen 1984). After estimation of the polychoric correlation matrix,1 a principal component analysis with a varimax rotation was applied to ascertain the factors. Before applying the varimax rotation, an oblique rotation (oblimin) was performed showing that the factors were not substantially correlated.2 Varimax rotation was chosen because it produces easily interpretable results.Th is procedure was fi rst followed to derive the dependent variable, attracting target groups, using the three items listed earlier. To create this variable, the factor scores of the single factor emanating from this analysis were saved (with an eigenvalue of 2.19930). Bartlett’s test of sphericity, with a p-value of .000, indicated no problem of intercorrelation between the items. A Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy was used to assess whether the factor analysis was good enough to proceed. Th e result of the KMO test was .608, which is above the frequently used critical threshold of .5 (Field 2009). Th e procedure and results of the factor analysis of the obstacles are discussed later.Common Source BiasBecause all of the measures in this article are based on respondents’ self-reports, there is a potential problem of common source bias.T o check for common source bias, Harman’s single-factor test (see, e.g., Andersson and Bateman 1997; Podsakoff and Organ 1986)—a factor analysis with all items underlying the independent and dependent variables—was conducted. Th is analysis resulted in four diff erent factors with an eigenvalue above 1 (the factor with the highest eigenvalue explained 36 percent of the variance). Th is indi-cates that common source bias is not likely to explain the research fi ndings (cf. Andersson and Bateman 1997).Regression AnalysisA multivariate hierarchical regression analysis was conducted with the dependent variable, results of place marketing in terms of attracting groups. Regression analysis was chosen because of its capacity to examine the relationships between multiple variables and the possibility to control for confound-ing infl uences on the relationship betweenobstacles and results of place marketing (cf.Graddy 1998). Th ree variables—the sizeof the municipality, the experience of therespondent, and the position of the respond-ent—were controlled for using hierarchicalregression analyses: the fi rst model includedonly the set of three control variables, andsubsequently the set of independent variableswas added.Th e fi rst control variable included was the size of the municipality, in order to control for the possibility that city marketers in larger cities may have to cope with more complex administrative conditions—for example, they have to deal with more departments within the municipality. Another possibility is that larger cities tend to be more active in place marketing than smaller ones, and this may also infl u-ence perceptions of obstacles or outcomes.Th e second control variable, the respondent’s experience in place mar-keting, was included to control for the possibility that the variance in outcomes is not explained by the obstacles but by incorrect esti-mations attributable to limited experience among our respondents. One could expect experience to have an eff ect on the evaluation of the outcomes by the respondents; respondents with more experience might evaluate the outcomes more correctly than those with less experience.Th e third control variable relates to the respondents’ diff erent posi-tions. Th ey work in diff erent organizations, and it was necessary to control for the possibility of their organizational affi liation infl uenc-ing their perception of obstacles or results. For example, respondents from tourism offi ces may have a diff erent outlook on administrative or political bottlenecks because they work in a diff erent organiza-tion than those from municipalities. Th ree diff erent positions match with the main categories of survey respondents: (1) public managers in municipalities (61.3 percent), (2) public managers in tourismoffi ces (24.8 percent), and (3) public managers in other organization (13.9 percent). Th e last group consisted largely of people working in independent associations responsible for place marketing. Dummies were created to measure this variable. Th e respondents working in municipalities were the reference group.Some Main Obstacles in Place MarketingTh e obstacles in place marketing were analyzed by fi rst exploring the most important obstacles according to practitioners in place market-ing, the survey respondents. Th e underlying dimensions of obstacles were analyzed by performing factor analysis. Th is section ends by stating three hypotheses on the relation between obstacles and out-comes, and this relation is explored in the following section.What Obstacles Do Practitioners Find Most Important? Presenting the median and the mode of each obstacle, table 2 sum-marizes how respondents perceive the obstacles. Th e median separates the population of respondents into two groups with an equal number of respondents: one group that leans more toward agreeing and one group that leans more toward disagreeing that this is an obstacle. A median of 4 indicates that half of the respondents strongly agree with the proposition that this is an obstacle in their municipality, whereasthe rest of the respondents do not agree thatthis is an obstacle (choosing a value of 3, 2,or 1). Th e higher the value of the median, themore people lean toward (strongly) agreeingthat this is an obstacle. Th e mode shows themost commonly chosen value.Table 2 shows that respondents (strongly)agree most often with three propositions:the budget for place marketing is too low,it is diffi cult to reach consensus within theRespondents (strongly) agreemost often with three proposi-tions: the budget for place mar-keting is too low, it is diffi cultto reach consensus within themunicipality, and place market-ing has insuffi cient impact onproduct development.Place Marketing as Governance Strategy: An Assessment of Obstacles in Place Marketing and Their Effects on Attracting Target Groups 511。
法国数学家拉格朗日著作《解析函数论》英文名
法国数学家拉格朗日著作《解析函数论》英文名Analysis of Functions by French mathematician LagrangeAnalysis of Functions, also known as Mémoire sur larésolution des équations numériques, is a groundbreaking work by French mathematician Joseph-Louis Lagrange. This seminal work, published in 1809, laid the foundation for the field of complex analysis and played a pivotal role in shaping modern mathematics.Lagrange's work in Analysis of Functions focused on the study of functions of a complex variable and their properties. He developed new methods for solving equations involving complex numbers, uncovering fundamental principles that would later become the basis of complex analysis. In particular, Lagrange's work on power series and their convergence properties was a major contribution to the understanding of complex functions.One of the key concepts introduced in Analysis of Functions is the concept of a holomorphic function, which is a complex function that is differentiable at every point in its domain. Lagrange's study of holomorphic functions and their propertieshelped lay the groundwork for the development of the theory of analytic functions, a central area of study in complex analysis.Analysis of Functions also includes Lagrange's work on the theory of residues, which are complex numbers associated with singularities of a complex function. Lagrange developed new techniques for calculating residues and applying them to the evaluation of complex integrals, a key tool in the study of complex functions.In addition to his mathematical contributions, Lagrange's Analysis of Functions had a significant impact on the development of mathematics as a whole. His work inspired future generations of mathematicians to explore the rich and diverse field of complex analysis, leading to further advancements in the study of functions of a complex variable.Overall, Analysis of Functions by Joseph-Louis Lagrange is a seminal work in the field of complex analysis that has had a lasting impact on the development of modern mathematics. Lagrange's innovative methods and profound insights continue to influence mathematicians to this day, making his work an essential reference for anyone studying the theory of functions of a complex variable.。
MonteCarlo(蒙特卡洛算法)算法
用Monte Carlo 计算定积分
考虑积分
I
x 1exdx,
0
0.
假定随机变量具有密度函数
fX (x) ex,
则
I E( X 1).
用Monte Carlo 计算定积分-
2
2
T
T
Monte Carlo 模拟连续过程的欧式 期权定价-
均匀分布
R=unidrnd(N)，-产生1到N间的均匀分布随 机数
R=unidrnd(N,n,m),产生1到N间的均匀分布 随机数矩阵
连续均匀分布
R=unifrnd(A,B) -产生(A,B)间的均匀分布随 机数
R=unifrnd(A,B,m,n)产生(A,B)间的均匀分布 随机数矩阵
Matlab 的随机数函数-
正态分布随机数
R=normrnd(mu,sigma) R=normrnd(mu,sigma,m) R=normrnd(mu,sigma,m,n)
特定分布随机数发生器 R=random(‘name’,A1,A2,A3,m,n)
例
a=random(‘Normal’,0,1,3,2) a=
基本思想和原理
基本思想：当所要求解的问题是某种事件出现 的概率，或者是某个随机变量的期望值时，它 们可以通过某种“试验”的方法，得到这种事 件出现的频率，或者这个随机变数的平均值， 并用它们作为问题的解。
原理：抓住事物运动的几何数量和几何特征， 利用数学方法来加以模拟，即进行一种数字模 拟实验。
实现从已知概率分布抽样
构造了概率模型以后， 按照这个概率分 布抽取随机变量 （或随机向量），这一 般可以直接由软件包调用，或抽取均匀 分布的随机数构造。这样，就成为实现 蒙特卡罗方法模拟实验的基本手段，这 也是蒙特卡罗方法被称为随机抽样的原 因。
Module-10-剑桥商务英语培训讲学
Be good at Motivating others Communicating Listening Solving problems Cooperating Dealing with Pretending observing
What is involve?
Traits Situational interaction Function Behavior Power Vision and values Charisma Intelligence
Laissez-Faire(free vein)
A free rein leader does not lead, but leaves the group entirely to itself as shown; such a leader allows maximum freedom to subordinates. They are given a free hand in deciding their own policies and methods.
the ice); Small talk, big payoff; Make friends; Strategy in a negotiation;
Chinese business manners
1. Why is Chinese business manners important?
2. How can we make a successful Chinese business ?
vacation at that time. 6. It was very kind of you to ask me, but I am afraid that
I will not be able to come. 7. We are so sorry that we cannot accept your kind
蒙特卡罗方法3.由巳知分布的随机抽样
蒙特卡罗方法
2019/2/7
2. 直接抽样方法
对于任意给定的分布函数 F(x) ，直接抽样方法如 下：
X n inf t , n 1,2,, N
F ( t ) n
其中，ξ1，ξ2，…，ξN为随机数序列。为方便起见， 将上式简化为：
X F inf t
F ( t )
若不加特殊说明，今后将总用这种类似的简化形 式表示，ξ总表示随机数。
蒙特卡罗方法
2019/2/7
证明
下面证明用前面介绍的方法所确定的随机变量序 列X1，X2，…，XN具有相同分布F(x)。
FX n ( x) P( X n x) P( inf t x)
F ( t ) n
P( n F ( x)) F ( x)
对于任意的n成立，因此随机变量序列X1，X2，…， XN具有相同分布F(x)。另外，由于随机数序列ξ1， ξ2，…，ξN是相互独立的，而直接抽样公式所确定的函 数是波雷尔（Borel）可测的，因此，由它所确定的X1， X2，…，XN也是相互独立的（[P.R.Halmos, Measure theory, N.Y.Von Nosrtand,1950]§45定理2）。
蒙特卡罗方法
2019/2/7
反应类型的确定方法为：产生一个随机数ξ
Pel 弹性散射
Pel Pin 非弹性散射
Pel Pin Pf 裂变
吸收
蒙特卡罗方法
2019/2/7
2) 连续型分布的直接抽样方法
对于连续型分布，如果分布函数 F(x) 的反函数 F－1(x)存在，则直接抽样方法是 ：
f (Xh) E P M h ( X ) h f (Xh) 1 h( X h )dX h M h( X ) M h
- 1、下载文档前请自行甄别文档内容的完整性，平台不提供额外的编辑、内容补充、找答案等附加服务。
- 2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
- 3、如文档侵犯您的权益，请联系客服反馈,我们会尽快为您处理(人工客服工作时间：9:00-18:30)。
,QFUHPHQWDO 0HVVDJHV 0LFUR .HUQHO 6HUYLFHV IRU )OH[LEOH DQG (IILFLHQW0DQDJHPHQW RI 5HSOLFDWHG 'DWD&DUORV 3pUH]'SW GH ,QIRUPiWLFD \ (OHFWUyQLFD 8QLYHUVLWDW GH 9DOqQFLD& 'RFWRU 0ROLQHU ( %XUMDVVRW 63$,1FDUORV SHUH]#XY HV*HUPiQ )DEUHJDW'SW GH ,QIRUPiWLFD8QLYHUVLWDW -DXPH ,&DPSXV GHO 5LX 6HF( &DVWHOOyQ 63$,1IDEUHJDW#LQI XML HV5 - 0DUWtQH] * 0DUWtQ,QVWLWXW GH 5REyWLFD8QLYHUVLWDW GH 9DOqQFLD3ROt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²ZKLFK ZH KDYH FDOOHG ³,QFUHPHQWDO 2SHUDWLRQV´ DIWHU LQFUHPHQWDO FKHFNSRLQWLQJ²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¶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¶V WR WKH UHFHLYHU¶V DGGUHVV VSDFH +RZHYHU WKH\ KDYH EHHQ GH VLJQHG WR XSGDWH GDWD FRSLHV PDLQWDLQHG E\ GLIIHUHQW SURF HVVHV )RU WKLV UHDVRQ WKH\ SUHVHQW WZR XQLTXH IHDWXUHV RQO\ WKRVH SDJHV IURP WKH YLUWXDO PHPRU\ UHJLRQ VSHFLILHG E\ WKH VHQGHU WKDW KDYH EHHQ PRGLILHG VLQFH WKH ODVW ,0 DUH WUDQVIHUUHG WR WKH UHFHLYHU DQG WKH GHVWLQDWLRQ RI WKH PHVVDJH LV DQ H[LVWLQJ YLUWXDO PHPRU\ UHJLRQ RI WKH UH FHLYHU¶V DGGUHVV VSDFH ZKLFK FRQWDLQV D FRS\ RI WKH VHQGHU¶V RQH %RWK VHQGHU 3 LQ )LJXUH DQG UHFHLYHU 3 PXVW UHVLGH LQ WKH VDPH QRGH DQG WKH HIIHFW RI VHQG LQJ DQ ,0 RQ WKHLU DGGUHVV VSDFHV LV GHSLFWHG LQ )LJXUH ,QLWLDOO\ 3 KROGV D EDFN XS RI 3 ¶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µPDS¶ ZKLFK VSHFLILHV WKH FRQWHQWV RI WKH RULJLQDO ,0 7KH PDS FRQWDLQV WKH VWDUWLQJ DGGUHVV DQG OHQJWK RI WKH VRXUFH UHJLRQ DV ZHOO DV D OLVW RI WKH SDJHV LQFOXGHG LQ WKH PHVVDJH7KH PDLQ REMHFWLYH RI 3,0V LV WR LPSURYH WKH PDQDJH PHQW RI ,0V E\ SURFHVVHV WKDW GR QRW KROG D FRS\ RI WKH VHQGHU¶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¶V DGGUHVV VSDFH MXVW DV LW KDSSHQHG ZLWK ,0V ,I WKH SDFNHG RSWLRQ LV XVHG WKHQ WKH ,0 LV UHFHLYHG DV D 3,0 DV LW KDV MXVW EHHQ GHVFULEHG EHIRUH )LQDOO\ ZKHQ WKH SDFNHG RSWLRQ LV XVHG ZKHQ ERWK VHQGLQJ DQG UHFHLYLQJ WKH ,0 WKH GDWD LV FRSLHG IURP WKH VHQGHU¶V WR WKH UHFHLYHU¶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³PRGLILHG ELW´ DV LW LV SURYLGHG E\ 008V EH H[WHQGHG LQWR D ELWPDS PDQDJHG E\ WKH PLFUR NHUQHO )URP WKLV SRLQW RQ ZH ZRXOG UHIHU WR WKLV ELWPDS DV ³WKH PRGL ILHG ELWPDS´ DIWHU ³PRGLILHG ELW´ (DFK HOHPHQW RI WKLV ELWPDS UHSUHVHQWV WKH VWDWH RI WKH SDJH PRGLILHG RU QRW UHJDUGLQJ RQH SRVVLEOH GHVWLQDWLRQ RI ,0V6LQFH XVHU OHYHO FRGH PXVW EH DEOH WR HVWDEOLVK ZKHQ HDFK FRS\ PXVW EH XSGDWHG ,QFUHPHQWDO 2SHUDWLRQV LQ FOXGH D SDUDPHWHU FDOOHG LQFUHPHQWDO GHVWLQDWLRQ 7KH PL FUR NHUQHO XVHV WKLV SDUDPHWHU WR ILQG RXW ZKLFK HOHPHQW RI WKH PRGLILHG ELWPDS VKRXOG EH FRQVLGHUHG WR HVWDEOLVK WKH VHW RI PRGLILHG SDJHV,PSOHPHQWDWLRQ ,VVXHV'HVSLWH WKH YHUVDWLOLW\ RI ,0V WKH LQWHUIDFH WKDW QHHGV WR EH SURYLGHG WR DSSOLFDWLRQV LV YHU\ FRPSDFW EHFDXVH ,0V DUH MXVW DQ H[WHQVLRQ WR WKH VWDQGDUG PHVVDJH SDVVLQJ LQWHU IDFH $SDUW IURP KHDGHU PDQDJHPHQW DQG LQFUHPHQWDO FRS\ WKH ZKROH VHW RI VHUYLFHV GHVFULEHG LQ VHFWLRQ KDV EHHQ LPSOHPHQWHG LQ WKH PLFUR NHUQHO EDVHG 967D RSHUDW LQJ V\VWHP > @ LQWHJUDWLQJ ,0V LQWR LWV RULJLQDO PHVVDJH SDVVLQJ LQWHUIDFH%DVLF ,QWHUIDFH7KH LQWHUIDFH FDQ EH EHWWHU GHVFULEHG LQ WHUPV RI D EDVLF PHVVDJH SDVVLQJ LQWHUIDFH FRPSULVLQJ RQO\ WKH VHUYLFHV UHFHLYH VUF PVJ DQG VHQG GHVW PVJ +HUH VUF DQG GVW DUH DQ\ YDOLG VRXUFH DQG GHVWLQDWLRQ LQ WKH H[LVWLQJ 26 DQG PVJ IRU WKH SXUSRVH RI H[FKDQJLQJ ,0V PXVW LQFOXGH WKH IROORZLQJ ILHOGV IODJV GDWD PDS DQG PVJLG 7KH PHDQLQJ RI WKLV ILHOGV GHSHQGV RQ WKH RSHUDWLRQ VHQG RU UHFHLYH DQG WKH W\SH RI ,0 EHLQJ H[FKDQJHG 7DEOHV DQG GH VFULEH WKHVH ILHOGV IRU ERWK RSHUDWLRQV7DEOH 0HVVDJH ILHOGV ZKHQ VHQGLQJ DQ ,07DEOH 0HVVDJH ILHOGV ZKHQ UHFHLYLQJ DQ ,0,QWHJUDWLRQ LQWR 967D¶V LQWHUIDFH967D¶V PHVVDJH SDVVLQJ LQWHUIDFH LV FOLHQW VHUYHU RUL HQWHG SURYLGLQJ WKUHH EDVLF VHUYLFHV PVJBVHQG PVJBUHFHLYH DQG PVJBUHSO\ &OLHQWV FDQ FRQWURO GDWD IORZ E\ PHDQV RI WKH 0B5($' IODJ :KHQ VHW WKH FOLHQW ZLOO UHFHLYH GDWD SURYLGHG E\ WKH VHUYHU LQ LWV FDOO WR PVJBUHSO\ RWKHUZLVH GDWD VHQW E\ WKH FOLHQW ZLOO EH WUDQVIHUUHG WR WKH VHUYHU DW PVJBUHFHLYH 6LQFH 967D¶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¶V PHVVDJH SDVVLQJ LQWHUIDFH WR VXSSRUW ,0V RQH LPSRUWDQW SUREOHP WKDW ZH KDYH DG GUHVVHG KDV EHHQ KRZ WR VKDUH WKH 008¶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¶ ZKLFK LV WKH WUDGLWLRQDO GLUW\ ELW PDLQWDLQHG WR IODJ LI WKH SDJH QHHGV WR EH ZULWWHQ WR GLVN LQ FDVH RI EHLQJ VWROHQ )LQDOO\ OHW XV FDOO ' WKH 008¶V GLUW\ ELW 7KH DOJRULWKP XVHG LV DV IROORZV:KHQ D SDJH LV WKH VRXUFH RI DQ LQFUHPHQWDO RSHUDWLRQ WKH NHUQHO,I ' LV VHW LW VHWV ERWK '¶ DQG HYHU\ ELW LQ 9 DQGWKHQ FOHDUV '&KHFNV 9 EHLQJ L WKH LQFUHPHQWDO GHVWLQDWLRQ IRUWKH RSHUDWLRQ DQG LI LW LV VHW LQFOXGHV WKH SDJH LQWKH RSHUDWLRQ 7KHQ LW FOHDUV 9:KHQ D SDJH LV WKH GHVWLQDWLRQ IRU DQ LQFUHPHQWDO RS HUDWLRQ WKH NHUQHO VHWV HYHU\ ELW LQ 9 EHFDXVH LW KDV FKDQJHG DQG WKHUHIRUH VKRXOG EH LQFOXGHG LQ WKH QH[W LQ FUHPHQWDO RSHUDWLRQ DQG FOHDUV ' '¶ VKRXOG EH VHW RQO\ LI LW VKRXOG EH ZULWWHQ WR GLVN LQ FDVH RI EHLQJ VWROHQ:KHQ WKH NHUQHO VWHDOV D SDJH&KHFNV ' ,I LW LV VHW WKHQ LW VHWV ERWK '¶ DQG HYHU\9 ELW 7KHQ FOHDUV '&KHFNV '¶ ,I LW LV VHW WKHQ ZULWHV WKH SDJH WR GLVNDQG FOHDUV '¶8VLQJ ,QFUHPHQWDO 2SHUDWLRQV,QFUHPHQWDO 2SHUDWLRQV DUH D XVHIXO WRRO IRU WKH LPSOH PHQWDWLRQ RI IDXOW WROHUDQW DSSOLFDWLRQV 7KH IROORZLQJ VXE VHFWLRQV VKRZ KRZ WKH\ FDQ EH XVHG LQ GLIIHUHQW VLWXDWLRQV HDVLQJ WKH GHVLJQ RI VHYHUDO IDXOW WROHUDQFH PHFKDQLVPV DQG LPSURYLQJ WKHLU SHUIRUPDQFH(UURU 'HWHFWLRQ DQG &RQILQHPHQW,W KDV EHHQ RXWOLQHG KRZ ,QFUHPHQWDO 2SHUDWLRQV RIIHU IHDWXUHV WKDW FDQ EH XWLOL]HG WR LQFUHDVH WKH UHOLDELOLW\ RI DSSOLFDWLRQV 7KHVH IHDWXUHV FDQ EH XVHG E\ XVHU PRGH OD\ HUV ZKHQ DSSOLFDWLRQV KDYH VSHFLDO GHSHQGDELOLW\ UHTXLUH PHQWV$V LW ZLOO EH VKRZQ WKH SRVVLELOLW\ RI UHFHLYLQJ WKH KHDGHU RI D PHVVDJH EHIRUH LWV FRQWHQWV DUH PDSSHG RQWR WKH UHFHLYHU¶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¶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µ0¶ PHDQLQJ PRGLILHG DQG µ ¶ QRW PRGLILHG UHJDUGLQJ ERWK LQFUHPHQWDO GHVWLQDWLRQV WR VLPSOLI\ WKH ILJXUH RQO\ WKHVH WZR HOHPHQWV RI WKH ³PRGLILHG YHFWRU´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¶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¶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¶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²ZLWK LWV RZQ YLUWXDO DGGUHVV UDQJH²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¶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。