A comparative study of two modeling approaches in neural networks
气升式环流反应器内气液两相流动CFD数值模拟的研究
internal baffle 1999
24.Couvert A.Bastoul D.Roustan M Prediction of liquid velocity and gas hold-up in rectangular air-lift reactors of different scales 2001
36.Warsito.Ohkawa M.Kawata N.Uchida S Cross-sectional distributions of gas and solid holdups in slurry bubble column investigated by ultrasonic computed tomography 1999
45.Iguchi M.Kondoh T.Morita Z I.Nakajima K Velocity and turbulence measurements in a cylindrical bath subject to
centric bottom gas injection 1995(02)
46.Deshpande N S.Joshi J B Simultaneous measurements of gas and liquid phase velocities and gas hold-up using laser Doppler velocimeter 1997
10.Visnovsky G.Claus J D.Merchuk J C Cultivation of Insect Cells in Bioreactors:Influence of Reactor Configuration and Superficial Velocity 2003
双光束激光焊接技术研究现状及应用_赵耀邦
90航空制造技术・2015 年第 23/24 期FORUM OF THE YEAR双光束激光焊接技术研究现状及应用 上海航天精密机械研究所 赵耀邦 成群林 李中权上海航天技术研究院 王玉平 阐述了双光束激光焊接技术特点及研究现状，对钛合金双光束激光焊接工艺进行研究，采用双光束激光焊能增强激光焊接的工艺适应性，减少焊接缺陷，焊接过程更加稳定。
试验研究表明，钛合金双光束激光焊焊缝成形良好，钛合金双光束激光焊接头强度与母材相当。
采用双光束激光焊接某钛合金薄壁结构件，满足设计要求的焊缝质量标准要求。
DOI:10.16080/j.issn1671-833x.2015.23/24.090Research Status and Application of Dual Beam Laser Welding Technology因激光焊的激光束光斑直径小，可精确控制焊接热输入，焊接线能量小，因此激光焊接变形和残余应力均很小，同时焊接热影响区窄、焊缝金属晶粒粗化倾向小。
对于钣金成形的薄壁结构件来说，由于成形精度难以控制，焊接装配间隙较大，且存在一定的错边，同时钛合金、铝合金激光焊的气孔倾向较大，给激光焊接技术在薄壁结构件的应用带来了困难。
为解决常规单光束激光焊存在的局限性，将单光束激光分离成两束激光，通过改变两束激光能量配比、光束间距、排布方式，对激光焊接温度场、流动场进行方便、灵活调节，改变匙孔的存在模式与熔池的流动方式，提高单激光焊的工艺适应性，为激光焊接工艺提供更加广阔的选择空间。
双光束激光焊接不仅拥有激光焊残余应力变形小、接头质量好等优点，且能较好解决常规激光焊接间隙适应性差、气孔倾向大、合金元素烧损严重等问题[1]。
TIG 焊、激光焊、双光束激光焊3种焊接工艺在焊接残余应力和变形、焊缝和热影响区宽度、间隙和错边适应性、气孔倾向、合金元素烧损和挥发等方面的比较如表1所示。
可看出双光束激光焊既保留了一般激光焊的优点，又克服了其固有缺陷。
火灾预防论文参考文献
火灾预防论文参考文献随着社会的发展，火灾已经成为了常见的事故之一，在生活中也随时存在着发生火灾的风险。
火灾不仅会给人们的生命财产造成重大的损失，更对社会的稳定和安全带来了威胁。
因此，对于火灾的预防和控制工作变得尤为重要。
本文将介绍一些值得参考的火灾预防论文参考文献以供读者借鉴。
一、火灾的危害性分析火灾的危害性十分大，可以造成严重人员伤亡和财产损失。
对于火灾的危害性进行分析，以便在预防和控制火灾时可以更全面地了解火灾的影响。
下面列举一些相关研究文献供参考：1. Andrzej Blusk, Katarzyna Topczewska, & Marcin Zawada (2019). Modeling fire consequences in the assessment of buildings’ safety on fire. Journal of Loss Prevention in the Process Industries, 57, 68-80.2. Gail Lopes, Paulo Santos, Ricardo Mateus, & Miguel Real (2019). Developing fire risk assessment models for largemixed-use offices. Fire Safety Journal, 107, 102864.3. Pengcheng Yang, Linna Wang, & Hongzhou Xu (2018).Multi-level fire risk assessment of residential buildings: A comparative study on difference and commonalities between two analysis methods. Fire Safety Journal, 99, 39-44.4. Kevin T. Roche, Yi Bao, & Jason D. Sutula (2018).Cost-Benefit Analysis of Fire Safety in Tall Buildings: Effect of Vertical Stratification on Smoke Spread and Evacuation. Fire Technology, 54(3), 907-934.以上文献重点分析了火灾对于建筑、人员、财产等方面的危害性，为火灾预防和控制提供了理论依据。
WINDMILLINGOFTURBOFANENGINE-TUDelft
WINDMILLING OF TURBOFAN ENGINECalculation of Performance Characteristics of a Turbofan Engine under Windmilling.P R I C E -I N D U C T I O N .C O Mponent simulations. It also models both continuous and discrete model systems. impacts di ff erent con fi gurations and preliminary dimensioning of equipment, mono-point and multi-point design, parametric studies, sensitivity analyses, customer deck generation, optimization studies, multi-fl uid models, maps han-dling, etc. The entire model is created by linking the di ff erent component models in a graphical user friendly interface. There are simple averaging techniques available to handle the 3D-0D-component data ex-change though the boundary conditions the whole engine model remain the same. While the boundary conditions of the 3-D simulations are automatically fed by the PROOSIS to the CFD software. The gure shows the schematic view of the DGEN 380 model in PROOSIS.PROOSIS primarily performs the simula-tions based on the thermodynamic gas turbine cycles using averaged variables describe the fl ow properties. The dif-tools are very accurate and result in much smoother maps. The similarity parameters used are scalars ffi ciency, mass fl ow and the speed calculated separately for rotating com-ponents. With the help of the steady ex-periment, PROOSIS able to simulate a large envelop conditions up to low pressure ratios MFT maps in order to run the steady state calculations. The graph in the gure shows steady state windmilling fuel is cut o ff and the pressure ratio drops and reaches a state where <1 but when the pressure ratio increases with higher mass fl ow rate the engine operates under normal condi-tion.The CFD simulations rst run for the fan blade alone large separations were seen near the tip. Adding the Outlet Guide Vane blades and using the mixing technique for OGV interface had little e ff ect on the ow across the fan.The turbine produced little work: roughly of the order of 10-15% of the design work. The high pressure compressor operates at portion is not much a ff ected compared to the hub part. Further, the lower part of the fan blade compresses a little while the up-per part expands the fl ow with an overall aerodynamic load of zero.CONCLUSION FUTURE WORKThe present work has been dealt with the behavior of the fan stage of a high bypass ratio turbofan engine-out conditions by reproducing windmilling operation in ground level test bed. The results dem-onstrate the challenges that arise in char-acterizing the fl ow due to the extremely low temperature and pressure variations. Work is in progress to complete the da-tabase with unsteady measurements to characterize the turbulent and unsteady components of the separated fl ow and provide a reference validation test case. Further work is constantly going on in im-proving the PROOSIS model. The extrapo-lation model is also being studied and improved to meet the new challenges. A detailed study of the MFT map methodol-ogy is also being studied for a step by step Figure 3. Stagnation pressure at the rotor exit Figure 1. Mach number on Fan (left) and OGV (right) of hub (top) and shroud (bottom)Figure 2. Steady state windmillingFigure 4. PROOSIS Model。
语言学导论第四章
XP rule:
XP
specifier the seldom very
(1) categories: ategories: a group of linguistic items which fulfill the same or similar functions in a particular language such as a sentence, sentence, a noun phrase or a verb. Kate likes classical music.(sentence) classical music (noun phrase) like (verb)
S Infl VP
Hale Waihona Puke Det Will theN train e
V arrive
Do insertion: insertion: Insert interrogative do into an empty Infl position. CP CP C NP N birds S Infl VP C NP S Infl VP V fly
three criteria for determining a word’s category
meaning inflection distribution
2.phrase categories: The category of phrases is determined by the word category around which the phrase is built. [NP a poor boy] boy] [VP run quickly, like music] [AP extremely cold] cold] [PP mainly about] about]
香港理工大学高分辨率的指纹（HRF）数据库_图像处理_科研数据集
⾹港理⼯⼤学⾼分辨率的指纹（HRF）数据库_图像处理_科研数据集⾹港理⼯⼤学⾼分辨率的指纹(HRF) 数据库(The Hong Kong Polytechnic University(PolyU)High-Resolution-Fingerprint (HRF)Database)数据介绍：Fingerprint is the most widely used biometric characteristic for personal identification because of its uniqueness and stability over time. Most of the existing automatic fingerprint recognition systems (AFRS) use the minutia features on fingerprints, i.e. the terminations and bifurcations of fingerprint ridges, for recognition. Although they can achieve good recognition accuracy and have been used in many civil applications, their performance still needs much improvement when a large population is involved or a high security level is required. One solution to enhancing the accuracy of AFRS is to employ more features on fingerprints other than only minutiae. Fingerprint additional features, such as pores, dots and incipient ridges (see Fig. 1 for examples), are routinely used by experts in manual latent fingerprint matching. Some of these additional features, e.g. pores, require high resolution fingerprint images to reliably capture them. Thanks to the distinctiveness of these fingerpr关键词：⾼分辨率的指纹,⾹港理⼯⼤学,UGC/CRC,High-Resolution-Fingerprint,PolyU,UGC/CRC,数据格式：IMAGE数据详细介绍：The Hong Kong Polytechnic University (PolyU)High-Resolution-Fingerprint (HRF) DatabaseOverview:Fingerprint is the most widely used biometric characteristic for personal identification because of its uniqueness and stability over time. Most of the existing automatic fingerprint recognition systems (AFRS) use the minutia features on fingerprints, i.e. the terminations and bifurcations of fingerprint ridges, for recognition. Although they can achieve good recognition accuracy and have been used in many civil applications, their performance still needs much improvement when a large population is involved or a high security level is required. One solution to enhancing the accuracy of AFRS is to employ more features on fingerprints other than only minutiae. Fingerprint additional features, such as pores, dots and incipient ridges (see Fig. 1 for examples), are routinely used by experts in manual latent fingerprint matching. Some of these additional features, e.g. pores, require high resolution fingerprint images to reliably capture them. Thanks to the distinctiveness of these fingerprint additional features and to the advent of high quality fingerprint imaging sensors, they have recently attracted increasing attention from researchers and practitioners working on AFRS.Our team in the Biometrics Research Centre (UGC/CRC) of the HongKong Polytechnic University has developed a high resolution fingerprintimaging device and has used it to constructed large-scale high resolutionfingerprint databases (HRF). We intend to publish our database to facilitate researchers designing effective and efficient algorithms for extracting andmatching fingerprint additional features.Fig. 1: Example additional features on fingerprints, pores, dots, and incipientridges.Description of the PolyU HRF Database:An optical fingerprint imaging device (see Fig. 2) has been built by our team.Its resolution is around 1,200dpi, and it can capture fingerprint images ofvarious sizes, e.g. 320*240 pixels and 640*480 pixels.(a) (b)Fig. 2: (a) The high resolution fingerprint imaging device we developed and (b) its inner structure.Two high resolution fingerprint image databases (denoted as DBI and DBII)have been set up by using the developed fingerprint imaging device. DBIconsists of a small training dataset and a large test dataset. The images of thesame finger in both databases were collected in two sessions which wereseparated by about two weeks. Each image is namedas “ID_S_X”.“ID” represents the identity of the person. “S” represents the session of the captured image. “X”represents the image number of each session. The following table gives the detailed information of the databases.We labeled the ground truth of sweat pores in 30 images selected from DBI. The central coordinates (row, col) of each pore were wrote into a text file (.txt). The ground truth of dots and incipients of 48 selected images were also offered. The central coordinates (row, col) of dots and two ends of each incipient were wrote into a text file (.txt). Here, the 48 selected images consists of 2 set of images ("SetIGroundTruthSampleimage" and "SetIIGroundTruthSampleimage") captured in two sessions. All of the original sample images and text files are contained in "Ground Truth.zip".Related Publication:1. Qijun Zhao, David Zhang, Lei Zhang, and Nan Luo, "AdaptiveFingerprint Pore Modeling and Extraction," Pattern Recognition, vol.43(8), pp. 2833-2844, 20102. Qijun Zhao, David Zhang, Lei Zhang, and Nan Luo, "High ResolutionFragmentary Fingerprint Alignment Using Pore-Valley Descriptors,"Pattern Recognition, vol. 43, pp. 1050-1061, 20103. Qijun Zhao, Lei Zhang, David Zhang, Nan Luo, and Jing Bao,“Adaptive Pore Model for Fingerprint Pore Extraction,” IAPR 19thInternational Conference on Pattern Recognition (ICPR2008), 20084. Qijun Zhao, Lei Zhang, David Zhang, and Nan Luo, “Direct PoreMatching for Fingerprint Recognition,” IAPR/IEEE 3rd InternationalConference on Biometrics (ICB2009), pp. 597-606, 20095. David Zhang, Feng Liu, Qijun Zhao, Guangming Lu, and Nan Luo,"Selecting a Reference High Resolution for Fingerprint RecognitionUsing Minutiae and Pores," IEEE Transactions on Instrumentation and Measurement, to appear6. Qijun Zhao, Feng Liu, Lei Zhang, and David Zhang, "A ComparativeStudy on Quality Assessment of High Resolution Fingerprint Images,"Proceedings of the IEEE International Conference on Image Processing (ICIP2010), Hong Kong, September 20107. Qijun Zhao, Feng Liu, Lei Zhang, and David Zhang, "Parallel versusHierarchical Fusion of Extended Fingerprint Features," Proceedings ofthe IAPR 20th International Conference on Pattern Recognition(ICPR'10), Istanbul, Turkey, August 20108. Feng Liu, Qijun Zhao, Lei Zhang, and David Zhang, "Fingerprint PoreMatching based on Sparse Representation," Proceedings of the IAPR20th International Conference on Pattern Recognition(ICPR'10), Istanbul, Turkey, August 20109. Q. Zhao, Lei Zhang, David Zhang, Wenyi Huang, and Jian Bai,“Curvature and Singularity Driven Diffusion for Oriented PatternEnhancement with Singular Points,” CVPR09. Proceedings of IEEEConference on Computer Vision and Pattern Recognition, pp. 1-7,Miami, Florida, USA, June 22-24 2009.The Announcement of the CopyrightAll rights of the PolyU HRF Database are reserved. The database is only available for research and noncommercial purposes. Commercial distribution or any act related to commercial use of this database is strictly prohibited. A clear acknowledgement should be made for any public work based on the PolyU HRF Database. A citation to "PolyU HRF Database, /doc/399f7c6fa45177232f60a2bf.html .hk/~biometrics/HRF/HRF.htm” and our related works must be added in the references. A soft copy of any released or public documents that use the PolyU HRF Database must be forwardedto: cslzhang@/doc/399f7c6fa45177232f60a2bf.html .hkDownloading Steps:Download “HRF DBI.zip”, “HRF DBII.zip”, or "Ground Truth.zip"to your local disk. Then, fill in the application forms. Send the application formto cslzhang@/doc/399f7c6fa45177232f60a2bf.html .hk. The successful applicants will receive the passwords for unzipping the files downloaded.HRF Databases:HRF DBI.zipHRF DBII.zipGround Truth.zipContact Information:Lei ZHANG, Associate ProfessorBiometric Research Centre (UGC/CRC)The Hong Kong Polytechnic UniversityHung Hom, Kowloon, Hong KongE-mail: cslzhang@/doc/399f7c6fa45177232f60a2bf.html .hk数据预览：点此下载完整数据集。
An Empirical Study of Machine Learning Algorithms Applied to Modeling Player Behavior in a
An Empirical Study of Machine Learning AlgorithmsApplied to Modeling Player Behavior in a “First Person Shooter” Video GamebyBenjamin Geisler(Advisor: Professor Jude Shavlik)A thesis submitted in partial fulfillment ofthe requirements for the degree ofMaster of Science(Computer Sciences)at theUNIVERSITY OF WISCONSIN - MADISONFall 20021 Abstract. Modern video games present many challenging applications for artificial intelligence. Agents must not only appear intelligent but must also be fun to play against. In the video game genre of the first person shooter an agent must mimic all the behaviors of a human soldier in a combat situation. The standard opponent in a “first person shooter” uses a finite-state machine and a series of hand coded rules. Drawbacks of this system include a high level of predictability of opponents and a large amount of work manually programming each rule. Modern advances in machine learning have enabled agents to accurately learn rules from a set of examples. By sampling data from an expert player we use these machine learning algorithms in a first person shooter. With this system in place, the programmer has less work when hand coding the combat rules and the learned behaviors are often more unpredictable and life-like than any hard-wired finite state machine. This thesis explores several popular machine learning algorithms and shows how these algorithms can be applied to the game. The empirical study includes decision trees, Naïve Bayes classifiers, neural networks, and neural networks trained using boosting and bagging methods. We show that a subset of AI behaviors can be learned by player modeling using machine learning techniques. Under this system we have successfully been able to learn the combat behaviors of an expert player and apply them to an agent in a modified version of the video game Soldier of Fortune 2. The following tasks were learned: speed of acceleration, direction of movement, direction of facing, and jumping. We evaluate both empirically and aesthetically which learner performed the best and make recommendations for the future. We also have created a system which uses these learned behaviors in a finite-state system within the game at real time.I.IntroductionSince their creation, the type of video games known as First Person Shooters (FPS) have been largely devoid of any sort of artificial intelligence. Games like Wolfenstein©, Doom© and Quake©present the player with enemies who mostly act as fodder. They walk towards the player, guns blazing, until they are dead. While this has its own merits, it has also become fairly boring. Recently developers began noticing this deficit and games such has Half-Life©, Counterstrike©, and UnReal© tournament have popped up. These games are successfully able to use expert-based systems and simple finite-state machines to give the illusion of an intelligent enemy (Linden 2001). As a result, these well established algorithms have helped game Artificial Intelligence (AI) advance in leaps and bounds in recent years. However, there is still much to be done. For the experienced player, the AI is never “good enough”. It needs to keep challenging the player, to adapt to the player and if possible learn from the player. Currently there is no such behavior in the realm of the First Person Shooter. Instead, a good player will learn the behavior of the enemy AI2and begin exploiting it. One of two things will happen at this point, either the player will abandon the game in boredom, or they will continue for entertainment value (after all, sometimes it’s fun to let off steam). Clearly games must account for both types of players.The answer to this conundrum may be found within the recent advances in academic artificial intelligence, especially machine learning. Machine learning is concerned with the question of how to build a computer program that can improve its performance on a task through experience. A task can be thought of as a function to be learned. It is the function’s responsibility to accept input and produce output based on the input parameters. For example, in making a decision to move forward or backwards in a FPS the input to the function is a list of variables describing the current environment of the game. The input vector will include data describing how many opponents are near the player, the players health level, and any other relevant information. The output is a classification of whether or not the player should move forward. The decisions made by a human player in a first person shooter will often be unexplainable except by example. That is, we can easily determine input/output pairs, but we cannot easily determine a concise relationship between the input and desired output. We can easily collect data of a good player. We can sample his actions and based on his performance in the game we can state with confidence that these are examples of correct decisions or examples of incorrect decisions. This process is what we investigated in this thesis.Machine Learning takes advantage of these examples. With these input/output pairs, a machine learning algorithm can adjust its internal structure to capture the relationships between the sample inputs and outputs. In this way the machine learner can produce a function which approximates the implicit relationships in the examples. Hidden amongst the examples will be many3 relationships and correlations. A human observer may be able to extract a few of these relationships but machine learning methods can be used to more extensively search out these relations.To use a machine learning algorithm it is first necessary to determine the features that are important to the task. It is also necessary to determine what we will attempt to learn from our examples. Once these features are determined, we can represent an input/output pair as a collection of feature settings and an indication of the desired classification of this input. For example, the input/output pair in our move forward/backward function would be a vector of environment data about the game and a decision to move forward or backward as output. In our hypothetical task we would then collect samples of data that fit the “move forward” classification and samples of data which fit the “move backward” classification. The combination of these two types of example classifications makes up the data set. The learning algorithm makes use of both types of examples when changing its internal structure. This form of learning is known as supervised learning, since we are acting as a teacher to the learning algorithm (Mitchell 1997). We specify which examples should be thought of as positive and which are negative. A supervised learning algorithm will split our data set further into a training set and a test set. The training set serves to allow the machine learner to update its internal settings based on only these examples. The test set is used to estimate how well the internal learned function will perform on previously unseen data (Mitchell 1997).There are many different machine learning algorithms to choose from. Some algorithms excel in certain domains while others do not perform as well. Instead of hastily picking a learning methodology it is preferable to frame the problem as much as possible. After the appropriate data4set is determined it will be necessary to gather plenty of samples, and try a few standard learning algorithms. However, “standard algorithms” to academics are often novel ideas to game developers. For example, things like ID3 trees, neural networks, and genetic algorithms have only recently been considered. Machine learning is almost non-existent in the realm of the FPS video game. This thesis will show that progress can be achieved by applying some recent machine learning algorithms to the task of learning player behavior in a First Person Shooter. This will enable game developers to insert some unpredictability to the agents in a video game. It will be shown that this application will also allow developers to model expert players with little rule specifications. This means the doors will be opened for many different behaviors for our agents. For example, it will be very easy to model the behavior of a sniper or heavy-weapons specialist without needing predetermined rules.With current trends in game development the timing couldn’t be better to introduce new learning algorithms. Computer gaming is now an six billion dollar a year industry (Savitz 2001). It is a competitive marketplace especially in terms of gameplay, which is something largely determined by the quality of a games’ AI routines. A game with a large variety of unique agent behaviors has a competitive advantage to a game with a few hand coded expert systems. In addition to being a preferred investment to developers it’s also a viable option on current hardware. Due to hardware advancements, processor cycles are easier to come by and memory is cheap. Neural networks in modern games would be unheard of in the late nineties. But now it’s not only possible on the PC, it’s already implemented in games such as “Black and White”™.5 II.The Problem and the Data SetIt is straight forward to take a well-defined game, throw in some control structure for some agents, and begin teaching them. But an FPS1 is not well defined. There has only been a spattering of attempts at doing so (Laird 2000). There are no defined tactics for a player of an FPS. The game is about battle, it is survival of the fittest; shoot or be shot. The more agents the player kills, the higher his score. But if the player is too gun happy, he might end up getting caught off guard and shot in the back. The penalties for death are severe; the player must wait thirty seconds before continuing the competition. Thirty seconds in this type of game is a significantly long time. In the meantime, the player’s opponent might have picked up a better gun, or a “health” pickup, or finished off a mutual opponent whom he’s been hunting since the beginning of the game (leaving the player with no points for the kill). Even expert players can’t fully quantify their tactics. They can tell you general tips. i.e., make sure to not stay in one place too long, do not rush at an opponent straight on, strafe to the left if one can. But there are so many exceptions to any one rule that it makes it difficult to hard code a rule system.To get a grip on all of this, the first step is to figure out what input the program has access to and what output might be useful. It might seem at first that an ideal FPS learning system would account for every action to take at any given situation. However, there will always be a need to allow some amount of programmer control in an FPS. The storyline and theme may demand specific actions at certain times. For example, it may be an interesting first encounter to set up a thug that always jumps out from behind a crate to scare the player. If we instead used player modeling to control the thug, the desired outcome may never happen. For this reason we want1 Although our research applies to any FPS, this research has used the FPS known as “Soldier of Fortune 2”™ (made by Raven Software) for its experiments.6control over what behaviors are learned from expert players and which are hand coded. Combat is a good place to use learned behavior because unpredicted (but sensible) actions are always preferred. Furthermore, there are only a handful of basic actions that are important during combat and these actions can be quantified. In this thesis we look at four of the basic actions: accelerate/decelerate, move forward/backward, face forward/backward, jump/don’t jump. The general behavior of a bot2 or player can be broken down into a sequence of moves (or move combinations). There will be many finer points to address once the sequence of moves is determined. For example, it’s expensive to evaluate where a bot can move every time frame3. Without proper calculations a bot may accidentally bump into a wall or walk off a cliff. Such pragmatics will be addressed latter in this thesis.What is needed to make a choice for any of these decisions? How does a player or agent know if it is time to move forward. This is at the heart of the problem. Typically these decisions are hand coded within some rule-based system. However, to achieve more lifelike and challenging characters, developers would like to model these decisions after an actual player and learn how these decisions are made. At this stage, the problem becomes a more traditional machine learning problem. A good set of features must be found to represent the environment of the game. This is the hard part of learning in a FPS, because even an expert gamer can not quantitize what it is that makes him good. Many of his decisions will be reflexive, and will not have an immediately tangible explanation behind them. With so much available data at any given time step, the selection of applicable features is very important.2 The AI agents in FPS video games are known as a “bots.”3 A frame is the smallest time step in our simulation.7 The Feature SetThe first step in modeling player behavior for an FPS video game is choosing the appropriate set of features for our learning algorithms. As mentioned, this thesis investigated four of the basic combat actions for an FPS: accelerate/decelerate, move direction, face direction and when to jump. These four actions will be the output portion of the data sample. We must now decide what information will be useful in making these four decisions. There is a large amount of available information but much of it is superfluous to learning player strategies. For example, knowing which 3D model file the player currently is using for his pistol is not useful to the learning algorithm. Perhaps the most important information in any combat scenario is to know where the agent’s enemies are. The more knowledge our agent has, the better his chances at surviving. Since information about location is so important, it is divided into several smaller features. Spatial data about enemy location will be determined by breaking the world up into four sectors around the player, as illustrated in Figure 1.Figure 1: Game sectors with player centered in the middle.8Sector information is relative to the player. For example, Sector 1 represents the area in front of the player. Actual world position is not as important to the player as is information about his immediate surroundings. Data about enemies and goals will be expressed in terms of sector coordinates. For example, the location of the closest enemy to the player will be indicated by a sector number. If the goal is due east of the player it will be considered in sector 2. A sector ends at 700 feet and is the maximum view distance of the player on the sample levels.Information about enemy location is extremely important to the player. Table 1 lists all the input and output features used for our learning algorithms. Of the twelve input features more than half of them report enemy statistics. The number of enemies in each sector will be recorded as will the health of the closest enemy, his distance from the player, and his sector location.In addition to enemy information, the player must continually be aware of his health level. For example, if a player has low health his actions might be significantly different than if he has full health. Also, the Soldier of Fortune 2™ game is not simply about surviving the longest without dying. Points are awarded to whomever can “capture the flag” and return it to their base. To an expert player this is significant motivation to change strategy during game play. The feature set includes information about the distance the player is from the closest goal and what sector contains that goal. Finally, we record the current move direction and direction the player is facing. This allows us to determine such tactics as retreat and advance.9 Table 1: Input Feature Set and Possible Outputs4If there are more than ten enemies the value is set to ten. The same upper bound holds for the other three sectors.Extracting the Data SetOnce the feature set is determined it is possible to run the game and collect some samples. For the purposes of this thesis, the outcome associated with any set of features will be one of the basic actions as described in the output section of Table 1. The feature set will be measured every other game frame (100 milliseconds). Every sample has a corresponding outcome that occurs 50 milliseconds later (the amount of time it takes to process a combination of player key-press events). This outcome will be some combination of the four basic actions:- Move front / back- Face front / back- Jump / do not jump- Accelerate / do not accelerateIII.Learning From the DataData was collected by sampling the decisions of an expert player. The level used was a standard “capture the flag” game with thirteen enemy agents. We created a special version of Soldier of Fortune 2™ to collect the available game data and translate it into our feature vectors. For example, when a game frame is sampled we can access locations of the nearest enemies within a certain radius, then using vector math we compute their location relative to the player. This technique is used for all the sector information features. The Move Direction and Face Direction features are computed by simply recording the current world Face Direction and Move Direction of the player. The output section of the feature vector is computed on the time step following the sample. We record whether or not the player accelerates, changes movement, changes facing, orjumps. This part of the feature vector represents the decision made by the player. Now that we have the input features as well as the decision, we have a complete feature vector. This feature vector is saved and the collection of samples becomes our training and testing sets used for applying the learning algorithms. We collected over ten thousand individual examples by observing one expert player for 50 minutes.A.Algorithm DescriptionsThe first set of experiments will test a few standard classifiers: decision trees (ID3), Naïve Bayes, and artificial neural networks (ANN) to judge how well they can learn from our training examples.ID3There are many decision tree algorithms to choose from. This thesis uses the ID3 decision tree learning algorithm (Quinlan 1986). Since the decision at any point in time is not mutually exclusive of the other decisions, this problem can be easily extended to four separate decision trees. One decision tree will represent each possible binary decision: Accelerate, Jump, Move Direction, and Face Direction.The training examples will be the feature vector at any time slice (the input), and the given action (the output). For each of the four decision trees the root node will be created and labeled with the feature that best classifies the examples. The standard information-gain formula was used. For the remaining examples, we recursively apply the ID3 algorithm and again find the best feature.After several test runs, it was evident that ID3 generates some very lengthy decision trees. To overcome this it was decided to implement the common “rule post-pruning” algorithm. Thisalgorithm uses a part of the training set as a tune set to estimate expected future accuracy. This accuracy information is then used to prune parts of the tree. In this thesis we pull out 10% of the training set to be used as a tune set for ID3 pruning. Rule-post-pruning can be summarized as follows:1.Grow the decision tree as normal from train data.2.Create one rule for each path from the root node to a leaf node.For example, assume we created this rule set:if (Enemy Sector = 1) ^ (Goal Sector = 2) ^ (Closest Goal = 2) then Jump = yesIf (Enemy Health = 1) ^ (Enemy Sector = 2) then Jump = no3.Calculate and save the tune set accuracy on the full set of rules.4.For each precondition in each rule:a.Consider removing the precondition from the rule. For example, in the rule setabove we would first consider removing (Enemy Sector = 1) from the first rule.On the next iteration we would consider removing (Goal Sector = 2). Afterconsidering the final precondition of the first rule, we move on to the firstprecondition of the second rule.b.Remove the precondition temporarily and score this new rule set using thetune data. Remember this score and the corresponding preconditionremoved.5.Calculate the best tune set accuracy over all the rule sets created in step 4.a.If the best accuracy is not greater than the previous best accuracy,stop pruning and use the saved rule set from the previous step.b.Otherwise if the best accuracy is greater than the previous best accuracy,remove the corresponding precondition which, when absent, led to the higheraccuracy. Save this rule set and return to step 4, considering this new rule setfor further pruning.6.Evaluate the final pruned rules by having them classify the test set.NAÏVE BAYESThe Naive Bayes Classifier is another type of learner (Mitchell 1997). It uses Baye’s theorem and assumes independence of feature values to estimate posterior probabilities. In this case it is anticipated that the features in and FPS are independent enough of each other to allow accurate classification. Classifications are made using the highest posterior probability. For example, if we have conditional independence then the probability of observing a conjunction of attributes Prob(x | c), where vector x contains the attributes and c is the result, is equal to the product of the probabilities for observing the individual attributes x I. In other words Prob(x | c)= {1,..,n} Prob(x i | c) where n is the number of attributes in x. Thankfully our features are mostly independent of each other because most of the features concern location data. For example, the number of enemies in front of the player is not dependent on the number of enemies behind him. But there are certain instances in game play when this is not the case. For example, high player-health levels may dictate that more enemies are placed near the goal, meaning the enemy sector features may be partially dependent on the health feature. However, this is a rarity because this is only triggered when teams need to be evened out5. In addition, for many practical tasks Naïve Bayes still leads to accurate prediction even when the independence assumption is violated (Domingos 1996).ANNAn artificial neural network (ANN) learns by using a training set to regress through the examples and learn in a non-linear manner. The basic back-propagation algorithm (with ten hidden units) was used in this project (Mitchell 1997).5This phenomena only occurs approximately one in every twenty games.The standard validation-set concept was used to avoid overfitting of the training set. Similar to the method of moving 10% of the training examples into a tuning set for ID3 pruning, 10% of the ANN training data will be moved into a tuning set to validate our learned function. After every five epochs of training we save the network weights at that time step and calculate the error on the validation set with these weights. If at any time the error rate is lower than the previous error rate from five epochs previous, training is stopped and the network weights of the previous validation step are used.B. Experimental MethodologyCollecting the DataBecause of the complexity of the FPS there are some implementational hurdles to collecting this data. Since the data was sampled at a rate of every 100 milliseconds, there will be many input sets that look exactly the same. If nothing has changed in 500 milliseconds, only one of these samples is recorded. However, if at least one of the features changed this sample will always be recorded. This is important, otherwise crucial feature/action pairs might be missed.Events with no corresponding action are discarded. A portion of these events may have proven useful: if no action is specified it could mean to stand still. However, rare events such as climbing ladders, opening doors, and going prone were not encoded. For this reason, it can not be assumed that the remaining non-classified situations dictate any particular action, so they were thrown out before learning began.66 Un-classified examples accounted for approximately 7% of the total collected examples.The game is run in multi-player mode with thirteen agents placed in a large multi-leveled environment. This environment includes places for every type of basic movement. This system for collecting data has been verified to work on any environment type with no necessary customization. However, the examples applied to the learning algorithms were always from the same environment. Data was collected over the course of several game sessions and combined into one massive data set of approximately 6000 examples. Each game was run by the same “expert” player, whose performance was fairly consistent.Estimating the Error Rate of the Learned ModelsAfter collecting the data each learning algorithm is applied in independently. After the algorithm has used the tuning set and training set, accuracy is estimated on the test set. The same test set of a thousand examples is used for all experiments. To get an idea of how much data will be needed in practice, each set of experiments was run on several different training set sizes.C. ResultsFigure 2 shows performance on the test set as a function of the size of the training set. The three basic algorithms exhibited satisfactory performance with error rates being low across all tasks. In general, the ANN always had lower error rates across all four tasks on the larger data sets sizes. On the smaller data sets, ID3 and Naïve Bayes sometimes achieved lower error rates than the artificial neural network (ANN). The lowest error rates for all four tasks were achieved by the ANN using a data set size of 5,000 examples. Figure 2 also shows a baseline error rate. The baseline illustrates the error rate achieved if the majority category is always guessed.For Face Direction the ANN began to achieve lower error rates than the other two algorithms at only 500 data set examples. The error rate on Face Direction task dropped all the way to 16%. The baseline hovered at about 40%.For Move Direction the ANN began to achieve lower error rates than the other two algorithms at 1,500 data set examples. The error rate on Move Direction dropped all the way to 7%. The baseline hovered at about 40%t.For Jump the artificial neural network began to achieve lower error rates than the other two algorithms at only 500 data set examples. The error rate on Jump dropped all the way to 14%. The baseline error rate hovered at about 3%. The error rate baseline for this decision is actually substantially lower than the error rate using ANN. This interesting result is due to few occurrences of positive Jump examples. In relation to the other recorded decisions, jumping is a rarity. Using “confusion matrices” we will investigate this issue in full in another section. Because of the nature of jumping in an FPS, it turns out a higher error rate on this task is acceptable. A complete discussion is presented later in this thesis.For Accelerate the artificial neural network began to achieve lower error rates than the other two algorithms at 1,500 data set examples. The ID3 error rate behavior was similar to the ANN error rate on this task. But the ANN was still more accurate; the error rate on this task dropped to 16%.。
A comparative study of sentences in Chinese and English.
A comparative study of Chinese and EnglishID number: 201855820Name: LI MINGMINProfessor:박순혁Abstruct: English belongs to the Indo-European languages, is a synthetic language, grammatical relations between words in a sentence depends mainly on the morphological changes of words and maintained by the consistent relation between subject predicate structure to express; and the Chinese are different, the Chinese belongs to the Sino-Tibetan language family, is an analytical language, grammatical relations between words not by morphological changes and maintained by the consistent relation between subject predicate structure to express, but through the function words and word order to express, so the Chinese passive sentence mainly with "is", "by", "have" to express. On the features of Chinese grammar, Chinese unlike English has various forms, so the Chinese passive sentence and the semantic relations are often more closely. Modern language school attaches great importance to the study of the passive sentence, and try to use the case grammar, valence theory, cognitive theory to the analysis of Chinese passive sentences, the theory has achieved significant results in the analysis of the characteristics of Chinese passive sentence in the.I am on the basis of previous studies, mainly on the syntax, semantics and pragmatics, further study of English and Chinese sentence similarities and differences between the historical evolution of the problem, to effectively solve the difficult points in learning Chinese and English passive sentences.Key words:Chinese and English difference grammar type comparison1.Introduction about Chinese and EnglishSubjects and predicates in English and Chinese have active and passive relations. Because English belongs to Indo European language and Chinese belongs to the Sino-Tibetan language, so it is different in terms of morphology and syntax. English is a comprehensive language. The grammatical relations between words and words in a sentence depend mainly on the morphological changes of the word itself.In Chinese and English, the form, development, and semantic, syntax, pragmatic of sentences have a great difference. Moreover, the comparative study of Chinese and English sentences is an important aspect of contrastive linguistics. There is a close relationship between Chinese and English historical grammar studies and Synchronic Grammar Studies. Chinese and English sentences have always been an important issue in the study of Chinese grammar. This issue has been actively discussed and studied by many scholars. Therefore, on the basis of drawing on the achievements of predecessors, I will continue to learn about it.2.The difference of cognitive science and typology between Chinese and English2.1The difference of cognitive scienceFrom the perspective of thinking mode, thinking mode is the way of thinking exercise, and it is the unity of thinking structure and thinking exercise. There is a great relationship between thinking and language. Language structure determines theprocess of thinking, and different national language form different ways of thinking. Therefore, the differences between Chinese and English form a different way ofthinking between Chinese and English nations.Westerners are accustomed to antagonisming between subject and object, and regard nature as an external thing. So they emphasize concepts, avoid generality, emphasize human rights, and advocate the opposition between man and nature, andman's request for nature. Therefore, the opposite thinking mode has been formed between subjects and objects. The Chinese are accustomed to unifying the subjectand object and regard man and nature as a whole. They emphasize human relations,ignore utensils, emphasize morality, and advocate the unity of man and nature, and harmony between man and nature. Therefore, the western philosophical view of the opposition between the subject and object and the unified Chinese subject and object determine that Chinese and English expressions have their own characteristics.2.1.1Chinese focus on parataxis, English focus on hypotaxisThe English speakers always use words and sentences structures to show meanings and realize the connection between words or sentences by logical relations between semantics or sentences. English lays stress on the tightness of formal structure in order to make the relationship between parts of sentences clear, and they usually use related words, parts of speech, and function words. However, Chinese expressions are relative random and implicit in form, and the conjunction of words or sentences is achieved by means of connectives or linguistic forms. Chinese does not attach importance to the tightness of formal structure, so they rarely use relevant words and sentences. In Chinese, the sentence coherence mainly depends on the semantic connection, and makes full use of the causal relation between the words and sentences to show the action sequence to cohere. Moreover, they hardly use related words but phrases and clauses. Connection between sentences in Chinese the inherent logic and formless relation to connect the combination of sentences, so in Chinese sentences, there is not exist any changes in words and often saves the connection between sentences.The connective elements in Chinese are not necessary, but express meanings of words and sentences in contexts. From the whole, we can understand the specific meaning and the relationship between sentences. It highlights the flexibility and conciseness of Chinese when use some logical connectives.Compare with Chinese passive sentences, English sentences are more compact and tight because of the various connectives which play a role in bonding the relationship within words. The relationship in English sentences needs the morphological changes of all forms of words to emphasize the syntactic forms and syntactic functions, and focus on the relationship between sentence meanings and logical forms. In the sentence structure components, with the verb as the core, the subject and the predicate form a spindle, andthe object and the complement extend along the main spindle, the attributive and the adverbial are branch line in the extension, and are connected with the main line with related words. Conjunctions (such as however, but, so, since, etc.), relation words (such as which, when, that, where, who, how, etc.), and preposition ( such as about, to , with, in , of , to, etc) are very important in English sentences. Try to compare these two sentences as below:1) Wo men wei pei yang si wei he xun lian er jiao jiao xue,zhe liang zhe dou neng kuo zhan si wei,zeng qiang xue sheng wei she hui zuo gong xian de neng li,zhe shi jiao yu de zhong ji mu di.We teach math for thinking and discipline, both of which expand the mind and increase the students’ability to function as a contributing individual in society: the ultimate goals of education.Chinese do not need to use connection words to express logical relations between sentences, but English must use word order to show conjunctions between sentences. Because Chinese words do not have any twists and turns, so when express ideas in Chinese sentences, there is not any change of grammatical forms, but use direct combination of various components within a sentence. In contrast, when English vocabulary changes, which leading to the characteristics of English syntactic changes mainly rely on words forms. Moreover, the change in the English language mainly focuses on verbs, such as predicates change according to the subjects, verbs have non-finite form. In English, numbers, case of nouns, tense and voice of the verbs; relative pronouns, relative adverbs and conjunctions between main clauses and subordinate clauses, play an important role to link up sentences. Thus, Chinese and English sentences are two different patterns.2.1.2Chinese focus on subject , English fo cus on objectThe Chinese people are accustomed to way of thinking of the unity between subject and object. The thinking mode of Chinese nationality is unitary subjective thinking mode. Its characters are the subject intervening in the object, and the object integrating into the subject, which posses a strong subjective initiative, but lack some objective scientific nature. This way of thinking is expressed in language: Chinese sentences are often used to personal subject, and often they are omitted when it is self-evident. For example, the Chinesesentence "I suddenly thought of a good idea", the focus of its thinking is "I", while the object "idea" is secondary.In contrast to Chinese, Westerners are accustomed to the way of thinking of the opposite of the subject and the object. The thinking mode of the Westerners has dual characters, subject and object,and the latter occupies the dominant position, because the traditional western thinking takes ‘nature’ as the cognitive object. This object thinking way makes Westerners focus on things rather than the subject themselves when observing things. Expressed in language, Objects are often used as subjects in English sentence, and generally cannot be omitted, and thus the passive form of predicate is produced. For example, the same meaning as the previous one, used to say in English: "A good idea suddenly struck me." Here object "a good idea" becomes the focus of thinking, and the thinking subject "me" is placed in a secondary position.2.1.3Chinese—‘scatter’ language, English—‘focus’ languageFrom the morphology change of sentence, western language is a "focus" language, while Chinese is "scattered " language. Generally speaking, Chinese often uses SVO sentence patterns, so many sentences have no focus, and clauses are arranged in logical order of development. While English is different, the predicate of English sentence must be a finite verb, which must be consistent with subject in person and number. If there are other verbs in a sentence, it must be consistent with the subject. Therefore, the predicate verb is the soul and focus in English sentences. Compared with Chinese, the way of English sentence-making is according to form chains, and the way of Chinese sentence-making is according to natural temporal and spatial order.In terms of the characteristics of the sentence, there are great differences in the way of observing things between Chinese and English. English people always pay attention to the individuality of thinking, the refinement and concretion of thinking, and grasp the most important information and focus from the complex and changeable things, and put the most important information in the main sentence and put subordinate information in the secondary position. In contrast, Chinese pay more attention to the generalization and comprehensiveness of the thinking, and are not good at catching the most importantinformation from the complex and changeable things, neglecting the relationship and priority between individuals.Try to compare:2) Lulled by the motion and soothed by the rippling music of the waves the babies soon fell asleep.Zhuan er qing qing yao dang,bo sheng chan chan yue er,hai zi men bu jiu jiu shui zhao le.In this example, ‘fell asleep’ is the predicate and the center in the whole sentence, and other verbs like ‘lull’ and ‘soothe’ use only –ing form. Therefore, it is very clear to see key information and sentence structure in English sentence, however, the Chinese one is just not clear and blurry, moreover, we cannot tell the difference in three sub-clauses above.2.2The difference of typology2.2.1 Chinese—analysis mode , English—integrate modeFrom the point of morphological classification, the world's languages can be classified into comprehensive and analytic languages. Analytic language means that the grammatical relation is not expressed through the morphological changes of the words themselves, but by means of function words and word order etc. Comprehensive language means that a language changes through the form of a word itself to express grammatical meaning, such as case, number, tense etc. According to this, we can infer that Chinese is a language based on analysis. In other words, grammatical relations are not expressed by the morphological changes of the words themselves, but by functional words and word order. On the other hand, English is a language developed from comprehensive to analytical language, that is, English mainly express grammatical meaning by means of its own change of forms.There are not any changes or turns in Chinese. Nouns will not change their form into plural, and verbs do not change their form to express the past tense, present tense and future tense. Modern English still holds some character of the comprehensive language, but it has the character of the analytic language: there are morphological changes; Word orders are more flexible than Chinese, but relatively fixed; Many of function words are used quite frequently. There are no morphologic changes in Chinese like Indo-European languages. But Chinese also has its own special features. There are two tense words exist in Chinesesentences: tense auxiliaries and time adverbs. There are three tense auxiliary words in Chinese: "着（zhe, v-ing）", "了（liao, v-ed）", "过（guo, have been）" and its collocation forms; And the usually used time adverbs are as follows: 总（zong, always）, 又（you, again）, 老（lao, again）, 再（zai, again）, and a large number of special time adverbs, such as 现在（xianzai, now）,过去（guoqu, past）and so on. However, English belongs to a comprehensive language, whose symbol is the inflection form of the predicate. An English verb itself has four forms: present, present participle, past and past participle, such as the four forms of the verb go: go, went, going, gone and so on. Then use these four forms can change into sixteen different tenses. At the same time, English verb forms must abide by various norms, and have strict requirements in terms of number, tense, and voice. The difference in tense between English and Chinese is mainly reflected in the diversification of verb forms. In Chinese sentences, the verb forms are more stable than English.2.2.2 Pronunciation differenceChinese phonemes are divided into two categories: vowels and consonants, and the same with English. A Chinese syllable is usually composed of one single vowel and one single consonant. The vowel (V) is the central component of a syllable, and consonant (C) is an auxiliary part of a syllable. There are four types of syllables in Chinese: V, CV, VC and CVC, and we can see that vowels are dominant ones when syllables are formed. In contrast, the syllable compositions in English are much more complicated. Besides the above four types, it can also be linked by three or four consonants, such as Aus-tra-li-a. Chinese is a tone language, and its tone feature endows Chinese with a cadence of musical beauty. Every Chinese character has four voices: high level, high rising, falling and rising, and falling. Tone is an indispensable part of syllables, and if the tone of Chinese character is different, although vowel consonants are same, meaning can be different. Such as: 韩（hán，family name）喊（hǎn，shout）汗（hàn, sweat）旱（hàn, dry）and so on. The rhythm of voice, degree, length, speed and the timbre of the speaker enables Chinese to express the rich thoughts and exquisite feelings accurately. Chinese sentences have intonation, which is generally displayed on the stressed words at the end of a sentence, but it cannot change the tone and is restricted by the tone.English is not a tone language, but an intonation language. The unit of the intonation is usually a sentence group or a clause, and the stress is usually at the end of a sentence. Intonation has not only abundant grammatical functions, but also can separate declarative sentences, interrogative sentences or disjunctive questions. Besides, they also have unique expressive effect. Different intonation can express different attitudes, tone, thoughts and feelings, and posses a melody beauty in English sentences. There is no tone in English words, but there is a stress. The regular arrangement of stress is an important phonetic feature of English. The difference of the position of stress indicates the different word class or meaning. The change of the position of sentence stress can strengthen the tone and highlight the contrast meaning.2.2.3 Vocabulary differenceChinese is one of the oldest languages in the world. Chinese vocabulary is rich and its meaning is abundant. With the development and change of language, Chinese vocabulary has developed from "monosyllabic" to "double syllable". The development trend of vocabulary makes it easier to clear the meaning of homophones, which makes the meaning of words become more accurate. Idioms are one of the most characteristic words in Chinese. They play a very important role in Chinese vocabulary. They are concise and comprehensive, which can be replaced and regulated, which giving Chinese expression more strong. There are many synonyms in Chinese. The nuances between synonyms not only enrich the expressive power of Chinese, but also embody the implicit beauty of word meanings. With the rapid expansion of foreign exchange communication, Chinese has absorbed many foreign words (such as computer, sofa, Coca-Cola, software, mouse and so on). They are not only products of foreign culture nut also embodiment of cultural integration.English is the most widely used language in the world. It absorbs Latin, Greek, French, German, Portuguese, Russian, and Chinese with its development and formation, which are the most important words in the world. It also has flexible methods of word formation, which can be used to form new words and new meanings by changing the inflection. The vocabulary of English is very rich, which makes the English ideographic both accurate and vivid, and can vividly express the delicate thoughts and feelings; the meaning of Englishwords are wide, the phenomenon of polysemy and many characters in one word are very common. The division of words in Chinese and English is different. There are two categories of Chinese notional words and function words. Notional words include nouns, adjectives, numerals, pronouns, verbs, adverbs, and quantifiers. Functional words have prepositions, conjunctions, interjection and auxiliary words. English is also divided into two categories: notional words and function words. Notional words are nouns, verbs, adjectives, adverbs, pronoun and numerals. There are no quantifiers in English; the function words include prepositions, conjunctions, auxiliary verbs and articles, and there are no articles in Chinese.2.2.4 Grammatical differenceThe difference in grammar between Chinese and English is mainly embodied in the following aspects: (I) there is no sign of the word class in Chinese, except for some limited suffixs such as子（zi）,员（yuan) and auxiliary words such as 的（de）, 地（de）, 得（de）. For example, "画（huà）" is a verb and a noun. We cannot know the feature only we see the whole sentence and analysis its structure. A Chinese verb itself does not have any grammatical form to express tense, voice and tone. There are articles in English and other Indo-European languages. Articles are used to mark nouns, and nouns are also the main parts of sentence grammar. After determining nouns, verbs can be basically determined in form accordingly. The changes in the tense of English verbs are expressed in the form of inflection, such as "He studies (studied, has studied, is studying)." Moreover, Adjectives pass through -ful, and adverbs are easier to identify by -ly. Second, nouns and pronouns in Chinese do not change in nominative or objective form. For example, "she loves me" and "I love her" as the subject of "she" and "I" are the same character to "I" and "she" as objects. In English, the nominative and the objective are the same and the forms are different. Third, subject and predicate of Chinese sentences only require the semantic consistency, without the corresponding change in form. In English, the subject predicate language is strict in terms of number, person and verb tense.For example:3) wo xihuan huahua, ta xihuan huahua, wo men da jia dou xihuan huahua.I like painting,she like painting,we all all like paintingI like drawing. She likes drawing. We all like drawing.3.ConclusionThe similarities between any two languages are universal and differences are relative. All human languages are symbolic and important tools of human life. The evolution of language has been developed from imperfection to perfection. Grasping of the differences accurately between Chinese and English helps us to understand the characteristics of these two languages better.。
临床医学硕士毕业论文参考文献范例[Word文档]
![临床医学硕士毕业论文参考文献范例[Word文档]](https://img.taocdn.com/s3/m/133d2c97d5d8d15abe23482fb4daa58da0111c06.png)
临床医学硕士毕业论文参考文献范例本文档格式为WORD,感谢你的阅读。
最新最全的学术论文期刊文献年终总结年终报告工作总结个人总结述职报告实习报告单位总结演讲稿临床医学硕士毕业论文参考文献范例一篇论文的是将论文在研究和写作中可参考或引证的主要文献资料，列于论文的末尾，以下是临床医学硕士毕业论文参考文献范例，欢迎阅读参考。
参考文献一：[1]VargasHA,Akin05ZhengJ，etal.ThevalueofMRimagingwhenthesiteofuterinecancerorig inisuncertain[J].Radiology,2011，258(3):785-792[2]HoriM，KimT，OnishiH，etalUterineTumors:Comparisonof3Dversus2DT2?weightedTur boSpin-EchoMRImagingat3.0T—InitialExperience1[J].Radiology,2011，258(1):154-163.[3]SubakLL，HricakH，PowellCB?etal.Cervicalcarcinoma:computedtomographyandm agneticresonanceimagingforpreoperativestaging[J].Obste tGynecol，1995，86(1):43-50.[4]MayrNA，YuhWT,ZhengJ,etal.Tumorsizeevaluatedbypelvicexaminatio ncomparedwith3-DMRquantitativeanalysisinthepredictionofoutcomeforcerv icalcancer[J].IntJRadiatOncolBiolPhys,1997，39(2):395-404[5]JensenRL，MumertML，GillespieDL，etal.Preoperativedynamiccontrast-enhancedMRIcorrelateswithmolecularmarkersofhypoxiaandv ascularityinspecificareasofintratumoralmicroenvironmen tandispredictiveofpatientoutcome[J].NeuroOncol,2014，16(2):280-291.[6]JansenJF，SchSderH，LeeNY,etal.Noninvasiveassessmentoftumormicroenvironmen tusingdynamiccontrast-enhancedmagneticresonanceimagingand18F-fluoromisonidazoIepositronemissiontomographyimaginginn ecknodalmetastases[J],IntJRadiatOncolBiolPhys,2010，77(5):1403-1410.[7]KimS，LoevnerLA,QuonH,etal.Predictionofresponsetochemoradiat iontherapyinsquamouscellcarcinomasoftheheadandneckusin gdynamiccontrast-enhancedMRimaging[J].AJNRAmJNeuroradiol,2010,31(2):262 -268.[8]JiaZ，GengD,XieT，etal.Quantitativeanalysisofneovascularpermeabilityingl iomabydynamiccontrast-enhancedMRimaging[J].JClinlNeurosci,2012，19(6):820-823.[9]YaoWW，ZhangH，DingB,etal.Rectalcancer:3Ddynamiccontrast-enhancedMRI;correlationwithmicrovasculardensityandclin icopathologicalfeatures[J].RadiolMed,2011，116(3):366-374.[10]OzdumanK，YildizE，DinnerA，ingintraoperativedynamiccontrast-enhancedT1-weightedMRItoidentifyresidualtumoringlioblastomasurger y[J].JNeurosurg，2014,120(1):60-66.[11]OfConnorJP，ToftsPS,MilesKA,etal.Dynamiccontrast-enhancedimagingtechniques:CTandMRI[J].BrJRadiol，2011，84(2):112-120.[12]JacksonA,JaysonGC,LiKL,etal.Reproducibilityofquant itativedynamiccontrast-enhancedMRIinnewlypresentingglioma[J].BrJRadiol，2003，76(903):153-162.[13]SevcencoS，PonholdL，JavorD，etal.Three-Tesladynamiccontrast-enhancedMRI:acriticalassessmentofitsusefordifferentiat ionofrenallesionsubtypes[J].WorldJUrol,2014，32(1):215-220.[14]RenJ5HuanY，WangH，etalDynamiccontrast-enhancedMRIofbenignprostatichyperplasiaandprostaticcar cinoma:correlationwithangiogenesis[J],ClinRadio,2008,6 3(2):153-159.[15]Shukla-DaveA,HricakH.RoleofMRIinprostatecancerdetection[J].NM RBiomed,2014,27(1):16-24.[16]deRooijM，HamoenEH,FuttererJJ，etal.AccuracyofMultiparametricMRIforProstateCancerDete ction:AMeta-Analysis[J].AJRAmJRoentgenol,2014，202(2):343-351.[17]WangXH，PengWJ5XinC,etal.[Valueofdynamiccontrast-enhancedMRIinassessmentofearlyresponsetoneoadjuvantche motherapyinbreastcancer][J].ZhonghuaZhongLiuZaZhi[Chin esejournalofoncology],2010,32(7):539-543.[18]TanSL,RahniatK，RozalliFI，etal.Differentiationbetweenbenignandmalignantbreastles ionsusingquantitativediffusion-weightedsequenceon3TMRI[J],ClinRadio，2014，69(1):63-71.[19]YoppAC，SchwartzLH，KemenyN,etal.Antiangiogenictherapyforprimarylivercance r:correlationofchangesindynamiccontrast-enhancedmagneticresonanceimagingwithtissuehypoxiamarke rsandclinicalresponse[J].AnnSurgOncol,2011，18(8):2192-2199.[20]ArmbrusterM,SourbronS，HaugA，etal.EvaluationofNeuroendocrineLiverMetastases:ACompar isonofDynamicContrast-EnhancedMagneticResonanceImaging紐dPositronEmissionTomography/ComputedTomography[J],Inve stRadiol,2014，49(1):7-14.[21]TappouniR，ElKadyRM，parisonoftheaccuracyofdiffusion-weightedimagingversusdynamiccontrastenhancementmagneti cresonanceimagingincharacterizingfocalliverlesions[J]. JComputAssistTomogr，2013，37(6):995-1001.[22]FolkmanJ.Tumorangiogenesis:therapeuticimplications [J].NEnglJMed，1971,285(21):1182-1186.[23]HawighorstH，KnapsteinPG,KnoppMV，etal.Cervicalcarcinoma:standardandpharmacokineticanaly sisoftime-intensitycurvesforassessmentoftumorangiogenesisandpati entsurvival[J].MAGMA,1999，8(1):55-62.[24]FranielT，HammB,HricakH.Dynamiccontrast-enhancedmagneticresonanceimagingandpharmacokineticmode lsinprostatecancer[J].EurRadiol,2011,21(3):616-626.[25]ToftsPS，BrixG，BuckleyDL，etal.Estimatingkineticparametersfromdynamiccontrast-enhancedT1-weightedMRIofadiffusibletracer:standardizedquantitiesa ndsymbols[J].JMagnResonImaging,1999，10(3):223-232.[26]AlonziR，PadhaniAR,AllenC.DynamiccontrastenhancedMRIinprostatec ancer[J].EurJRadiol,2007，63(3):335-350.[27]EllingsenC，EgelandTA，GulliksrudK,etal.Assessmentofhypoxiainhumancervicalcar cinomaxenograftsbydynamiccontrast-enhancedmagneticresonanceimaging[J].IntJRadiatOncolBio lPhys?2009,73(3):838-845.[28]SeeberL，HorreeN，VooijsMA,etal.Theroleofhypoxiainduciblefactor-1alphaingynecologicalcancer[J],CritRevOncolHematol，2011，78(3):173-184.[29]PurdieTQHendersonE，LeeTY.FunctionalCTimagingofangiogenesisinrabbitVX2soft -tissuetumour[J].PhysMedBiol,2001,46(12):3161-3175.[30]FidlerIJ.Angiogenicheterogeneity:regulationofneopl asticangiogenesisbytheorganmicroenvironment[J].JNatlCa ncerInst,2001，93(14):1040-1041.[31]田丽，刘立志，范卫君.鼻咽癌MRI动态增强参数与微血管密度和血管内皮生长因子表达的相关性研究[J].中山大学学报(医学科学版),2009,30(3):336-340.[32]于德新，马祥兴，魏华刚，等.肝细胞癌微血管及成熟血管对MR增强表现特征的影响[J].实用放射学杂志,2009，25(3):351-355.[33]DonaldsonSB，WestCM,DavidsonSE，etal.AcomparisonoftracerkineticmodelsforT1-weighteddynamiccontrast-enhancedMRI:Applicationincarcinomaofthecervix[J].MagnR esonMed,2010，63(3):691-700.[34]VeeravaguA,HouLC，HsuAR，etal.Thetemporalcorrelationofdynamiccontrast-enhanced,magneticresonanceimagingwithtumorangiogenesis inamurineglioblastomamodel[J].NeurolRes，2008，30(9):952-959.[35]HarrerJU,ParkerGJ，HaroonHA,parativestudyofmethodsfordeterminingv ascularpermeabilityandbloodvolumeinhumangliomas[J].JMa gnResonImaging，2004，20(5):748-757.[36]ZaharchukG.TheoreticalbasisofhemodynamicMRimagingt echniquestomeasurecerebralbloodvolume,cerebralbloodflo w,andpermeability[J].AJNRAmJNeuroradiol,2007，28(10):1850-1858,[37]McDonaldDM,BalukP.Significanceofbloodvesselleakine ssincancer[J].CancerRes，2002，62(18):5381-5385.[38]McDonaldDM,ChoykePL.Imagingofangiogenesis:frommicr oscopetoclinic[J].NatMed,2003，9(6):713-725.[39]MichelCC，CurryFE.Microvascularpermeability[J].PhysiolRev,1999，79(3):703-761.[40]YangC，StadlerWM,KarczrnarGS，parisonofquantitativeparametersincervixcancerm easuredbydynamiccontrast-enhancedMRIandCT[J].MagnResonMed,2010，63(6):1601-1609.参考文献二：[1]吴海树.齿科激光煌接理论与实践[M].大连：大连海事大学出版社，2002:49-50.[2]J.A.Atoui,D.N.B.Felipucci,V.O.Pagnano,etal.Tensilea ndflexuralstrengthofcommerciallypuretitaniumsubmittedt olaserandtungsteninertgaswelds[J].Braziliandentaljourn al，2013，24(6):630-634.[3]R.Nomoto,Y.Takayama，F.Tsuchida，etal.Non-destructivethree-dimensionalevaluationofporesatdifferentweldedjointsand theireffectsonjointstrength[J],Dentalmaterials,2010，26(12):246-252.[4]F.Caiazzo,V.Alfieri，G,Corrado,etal.InvestigationandOptimizationofLaserWeld ingofTi-6Al-4VTitaniumAlloyPlates[J].JournalofManufacturingScienceandEngineering-transactionsoftheASME，2013，135(6):452-463.[5]S.Michio，Y.Satoshi,T.Misao,etal.Influenceofirradiationcondition sonthedeformationofpuretitaniumframesinlaserwelding[J] ,DentalMaterialsJournal,2009，28(2):243-247.[6]H.Kikuchi,T.Kurotani,M.Kaketani,etal.Effectoflaseri rradiationconditionsonthelaserweldingstrengthofcobalt-chromiumandgoldalloys.[J]JournalofOralScience，2011，53(3):301-305.[7]X.J.Cao，A.S.H.Kabir，P.Wanjara,etal.GlobalandLocalMechanicalPropertiesofAut ogenouslyLaserWeldedTi-6A1-4V[J].MetallurgicalandMaterialsTransactionA-PhysicalMetallurgyandMaterialsScience，2014，45A(3):1258-1272.[8]S.Viritpon,Y.Takayuki,K.Equo，parativestudyontorsionalstrength，ductilityandfracturecharacteristicsoflaser-weldeda+PTi-6Al-7Nballoy,CPTitaniumandCo-Cralloydentalcastings[J].DentalMaterials,2008，24(6):839-845.[9]J.Liu，V.Ventzke，P.Staron,etal.EffectofPost-weldHeatTreatmentonMicrostructureandMechanicalProperti esofLaserBeamWeldedTiAl-basedAlloy[J].MetallurgicalandMaterialsTransactionA-PhysicalMetallurgyandMaterialsScience,2014，45A(l):16-28.[10]K.Hisaji,K.Tomoko，K.Masahiro，etal.Effectoflaserirradiationconditionsonthelaserweldi ngstrengthofcobalt-chromiumandgoldalloys[J].JournalofOralScience,2011,53( 3):301-305.[11]崔广，宋应亮.不同杆圈接触形态对辉接面机械性能影响的研究[J].中国美容医学，2012,21(12):2227-2229.[12]R.R.Wang,C.T.Chang.Thermalmodelingoflaserweldingfo rtitaniumdentalrestorations[J]JProsthetDent,1998，79(3):335-342.[13]J.M.C.NunezPantoja，A.P.Farina，L.G.Vaz，etal.Fatiguestrength:effectofweldingtypeandjointdesign executedinTi-6A1_4Vstructures[J].Gerodontology,2012,29(2):el005-el010.[14]解光明，王前文,夏荣.激光辉接钛及钛合金腐蚀疲劳强度的研究[J].实用口腔医学杂志，2009，25(4):463-464.[15]J.M.C.NunezPantoja，L.G.Vaz,M.A.A.Nobilo，etal.Effectsoflaserweldjointopeningsizeonfatiguestreng thofTi-6A1-4Vstructureswithseveraldiameters[J],JournalofOralRehab ilitation,2011，38(3):196-201.[16]LA.Orsi,L.B.Raimundo,O.L.Bezzon,etal.Evaluationofa nodicbehaviorofcommerciallypuretitaniumintungsteninert gasandlaserwelds[J].JournalofProsthodontics，2011，20(8):628-631.[17]卢军霞，曹风华，郭天文等.激光燥和脉冲氩辉对铁耐腐烛性的影响[J].实用口腔医学杂志，2000，17(2):105-108.[18]方洪渊.揮接结构学[M].北京：机械工业出版社，2008:109-127.[19]G.Bussu,P.E.Irving.Theroleofresidualstressandheata ffectedzonepropertiesonfatiguecrackpropagationinfricti onstirwelded2024-T351aluminiumjoints[J].InternationalJournalofFatigue,2 003，25(1):77-88.[20]宋天民.燥接残余应力的产生与消除[M].北京：中国石化出版社，2005:13-17.[21]I.Watanabe,C.Jennifer,Y.Chiu.Dimensionalchangeofla ser-weldedgoldalloyinducedbyheattreatment[J].BasicScienceR esearch,2007，16(5):365-369.[22]I.Watanabe,A.P.Beuson,k.Nguyen.Effectofheattreatme ntonjointpropertiesoflaser-weldedAg-Au-Cu-PdandCo-Cralloys[J].JProsthedont,2005，14(3):170-174,[23]I.Watanabe,J.Liu,M.Atsuta,etal.Effectsofheattreatm entsonmechanicalstrengthoflaser-weldedequi-atomicAuCu-6at%alloy[J].JDentRes，2001,80(9):1813-1817.[24]X.S.Liu,H.Y.Fang，S.D.Ji，etal.ControlofTitaniumAlloyThinPlateWeldingDistortionb yTrailingPeening[J].Mater.Sci.Technol，2003,19(1):184-186.[25]JJ.Xu，L.J.Chen，J.H.Wang.Predictionofweldingdistortioninmultipassgirth -buttweldedpipesofdifferentwallthickness[J].TheInternat ionalJournalofAdvancedManufacturingTechnology,2008，35(9-10):987-993.[26]A.S.Munsi,A.J.Waddell,C.A.Walker.Vibratorystressre lief-aninvestigationofthetorsionalstresseffectinweldedshaft s[J].JournalofstrainAnalysis,2001，36(5):453-463.[27]李占明，朱有利，王侃.超声波冲击处理对2A12招合金揮接接头组织的影响[J].金属热处理，2008,33(7):53-56.[28]R.T.Huang,W.L.Huang,R,H.Huang,etal.Effectsofmicros tructuresonthenotchtensilefracturefeatureofheat-treatedTi-6Al-6V-2Snalloy[J].MaterialsScienceandEngineering:A，2014，595:297-305.[29]T.Novoselova,S.Malinov,W.Sha.Experimentalstudyofth eeffectsofheattreatmentonmicrostructureandgrainsizeofa gammaTiAlalloy[J].Intermetallics,2003，11(5):491-499.[30]S.K.Nirmal,S.Rakesh,S.S.Vishal.CryogenicTreatmento fToolMaterials:AReview[J].MaterialsandManufacturingPro cesses，2010，25(10):1077-1100.[31]V.Firouzdor，E.Nejati,F.Khomamizadeh.Effectofdeepcryogenictreatment onwearresistanceandtoollifeofM2HSSdrill[J].JournalofMa terialsProcessingTechnology，2008，206(1-3):467-472.[32]J.W.Kim，J.A.Griggs,J.D.Regan,etal.Effectofcryogenictreatmenton nickel-titaniumendodonticinstruments[J].InternationalEndodont icJournal，2005,38(6):364-371.[33]S.Yan，Y.Zhang,S.G.Shi，etal.Effectoftwokindsofannealingmethodsonthemechanicalpropertiesofdentaltitaniumcastings[J].Journalofpractic alstomatology，2010，26(4):465-468.[34]闫澍，张玉梅，赵永庆等.热处理温度对牙科铸造纯铁拉伸性能的影响[J].实用口腔医学杂志，2008,24(1):74-77.[35]A.Ebihara,Y.Yahata,K.Miyara，etal.Heattreatmentofnickel-titaniumrotaryendodonticinstruments:effectsonbendingpr opertiesandshapingabilities[J].InternationalEndodontic Joximal,2011，44(9):843-849.[36]OA.Cuoghi，G.RKasbergen,P.H.Santos,etal.Effectofheattreatmentonst ainlesssteelorthodonticwires[J].BrazOralRes.2011，25(2):128-134.[37]S.S.D.Rocha，G.L.Adabo，L.G.Vaz，etal.Effectofthermaltreatmentontensilestrengthofcommer ciallycastpuretitaniumandTi-6AI-4Valloys[J],JMaterSciMaterMed,2005，16(8):59-66.[38]梁锐英，赵艳萍，温黎明等.激光燥接新型钴络合金的力学性能[J].中国组织工程研究与临床修复，2011，15(38):7135-7138.[39]刘亚俊，李勇，曾志新等.深冷处理提高硬质合金刀片耐磨性能的机理研究[J].工具技术，2001，35⑶:19-20.[40]D.N.Collins.Deepcryogenictreatmentoftoolsteels[J]. HeatTreatmentofMetals，1996，23(2):40-42.阅读相关文档:金融学硕士毕业论文参考文献范例管理学硕士毕业论文参考文献行政管理学毕业论文参考文献经济学硕士毕业论文参考文献范例经济学毕业论文参考文献格式管理学硕士毕业论文参考文献范例法律毕业论文参考文献格式范本经济学硕士毕业论文参考文献经济管理毕业论文参考文献经济硕士论文参考文献就业论文参考文献产业经济论文参考文献动画设计论文的参考文献范例关于动画设计专业论文的参考文献动漫设计论文参考文献三维动画论文参考文献动漫设计专业论文参考文献动画设计毕业论文参考文献动画设计专业论文参考文献推荐动画设计专业论文参考文献英语论文参考文献格式模最新最全【办公文献】【心理学】【毕业论文】【学术论文】【总结报告】【演讲致辞】【领导讲话】【心得体会】【党建材料】【常用范文】【分析报告】【应用文档】免费阅读下载*本文若侵犯了您的权益，请留言。
复杂抽样推断方法体系的比较研究(81)
复杂抽样推断方法体系的比较研究1贺本岚1 金勇进2（1.中国人民大学统计学院 2. 教育部重点研究基地中国人民大学应用统计科学研究中心北京100872）摘要：对复杂样本进行推断通常有两种体系：传统的基于随机化理论的统计推断和基于模型的统计推断。
传统的抽样理论以随机化理论为基础，将总体取值视为固定，随机性仅体现在样本的选取上，对总体的推断依赖于抽样设计。
该方法在大样本情况下具有稳健估计量，但在小样本、缺失数据等情况下失效。
基于模型的抽样推断认为总体是超总体模型中抽取的一个随机样本，对总体的推断取决于模型的建立，但在不可忽略抽样设计下估计量是有偏估计。
本文在对这两类推断方法核心内容进行简要分析的基础上，提出抽样设计辅助的模型推断，并指出该方法在复杂抽样中有重要的应用价值。
关键词：基于设计基于模型抽样设计复杂抽样推断A Comparative Study of Complex Sampling Inference SystemsBen-lan He Yong-jin JinAbstract: There are usually two inference systems for complex sample: traditional statistical inference and model-based statistical inference. Traditional sampling theory based on randomization theory believed that the values of variables on population units are fixed and the randomness embodies in sample selection. Its inference for population depends on sampling design. Estimators of this method are robust when sample size is large, but inefficiency when sample size is small and there are missing data. Another deduction based on model thinks that the finite population is a random sample drawn from a superpopulation. Inference for population depends on modeling, but estimators of this inference system are biased under nonignorable sample scheme. Based on the analysis of the core contents of the two methods, this paper proposes sampling scheme-assisted and model-based inference, and points out that this method has important application value in complex sampling.Keywords:Design-Based Model-Based Sampling Scheme Complex Sampling Inference1本文得到教育部重点研究基地重大项目“复杂抽样中的模型方法研究”（项目编号：10JJD790036）、国家社科基金项目“普查数据质量的事后抽查理论及其应用研究”（项目编号：11BTJ009）资助。
新世纪研究生英语b课后习题及答案考试复习
新世纪研究生英语课后习题及答案Unit11. The representative众议员presented to the committee a petition 请愿书signed by 1;200 electors asking for a thorough probe into the financial scandal surrounding the candidate for the regional legislature.2. The landlady fired the servant who pocketed独吞household funds for his own use.3. After exploring for more than half a year; the ecologist felta great from thrill兴奋 spotting therare species at the foot of the mountain.4. “Better late than never” is a platitude陈词滥调that is very familiar to most English speakers.5. The couple had been trying to satisfy all the needs of their only youngster; who had been importuning 胡搅蛮缠for more pocket moneny all the time.6. The unruly不守规矩的 crowd became even more boisterous喧闹的when the negotiator tried to quiet them.7. The belief that one should work hard and be honest is deeply ingrained根深蒂固 in our culture.8. The company was financially cornered把…逼入绝境and almost went into bankruptcy破产during last year’s economic recession.9. The employees tried to avoid every risk that might incur导致her displeasure during the firstmonth after her divorce.10. In order to prevent a sudden outbreak of hostilities; we must make no provocative刺激的;挑拨的move.11. The author of the book made a comparative study of courtesy 礼貌;礼节 between the two different cultrues. etiquette 礼貌;礼节12. After long night of revelry狂欢at the New Year’s Eveparty ;most of the guests were drunk喝醉的.looped 喝醉的13. Since the beginning of last month; she has suffered now and then from a sharp and unexplained 不明原因的headache for which the doctors are trying to find a cure. unaccountable不明原因的14. The security guard spoke angrily and seriously气愤的斥责 to the group of noisy boys acting up in the garden of the hotel. told off 斥责15. Compared with all his peers; the young engineer had been noticeably显着地successful in that field.conspicuously显着地16. Many men tend to regard daily housework such as cooking; cleaning and laundry to be an irritating使人烦恼的nuisance. pesky使人烦恼的17. The psychiatrist encouraged his patient to release释放all his resentment at the way life had treated him.let loose释放;放松18. In this lively生动的 romantic historical novel; the author narrates a riveting引人入胜的 love story between two great figures of the 19th century. racy 生动的19. She was depressed by the gloomy阴郁的;沮丧的; fog-bound scene on the waterfront and felt even more lonely.murky 阴郁的;沮丧的20. The chairman advised both parties to calm down and be rational because thy had reached a sensitive敏感的stage of their business negotiation.ticklish棘手的;敏感的翻译：To invite eminent persons to help make advertisements should be regarded as one of the best advertising strategies and could; of course; produce a spectacularpowerful VIP effect; privided that those celebrities are perfectly willing to accept the invitation and; more importantly; the products to be advertised are genuine and of fair prices. Sometimes; while a commodity is of inferior quality; the advertisement is full of words lavishing praise on it; if a celebrity shows up as an image agent for such a product; the advertisement could; if any; be temporarily successful before it turns the brand of the product in question notorious and; more disastrously; ruins the reputation of the eminent person thereafter.So; the famous are well advised to think more than twice before they agree to appear on the commercial.Unit21. Italian neo-realist新实在论者cinema focused on working—class subjects; using amateur actors and filming on location; rather than in the studio.2. Despite his wartime atrocities;暴行Hitler is still revered by neo-Nazi新纳粹主义的 in many countries.3.Neo-fascist新法西斯主义者groups are an increasing threat to ethnic minorities across Europe.4. By neocolonialism新殖民主义we mean the practice of granting a sort of independence with the concealed intention of making a liberated country a client—state.5. The tutor tried hard to dissuade阻止 students from using neologisms新词in formal writing6. He is a neo-Darwinist新进化论者animals is principally determined by natural selection; and that acquired characteristics cannot be inherited.7. The architect traveled around the whole of Europe seeking Palladio’s wonderful 16th-century neoclassical新古典主义的buildings.8. In view of the still unknown physical risks that cloning might impose on neonates婴儿appropriate.9. The challenge for the future lies in fostering培养 conditions in cities that will be favorable to economic growth.10. The old lady reaved赞扬 about her grandchildren for about an hour.11. Gene therapy-----the actual correction or replacement of defective gene sequences in the embryo or the adult ------is the holy grail最受欢迎的 of genetic medicine.12. Through genetic engineering; will scientists be tampering with the genetic diversity that has been the mainstay支柱 of human survival in the past13. A news release by the Chinese Ministry of Personnel printed ina few leading newspapers and on news websites caught the attention of thousands of twenty-somethings人到二十in October.14. For many reasons; the vast majority of heterosexuals异性的still prefer the “old—fashioned”; sexual way of producing children. No other method better expresses the loving union of a man and a woman seeking to make a baby.15. It is absolutely mandatory强制的 that every driver of a vehicle must pass a driver’s l icense test.16. In today’s lightning—speed world many singles are embracing 拥抱;接受 the newest trend in relationship ----- speeddating-------to size up a potential love interest.17. As a young man; he spent a lot of time meditating冥想upon the purpose of life.18. She really enjoyed the film because it was an fusion融合 of history and contemporary events.翻译：Nowadays in the city’s tonier residential districts there are peple named as singles; who are usually young; rich and tech-savvy professionals and choose independently their own lifesyles. The number of singles has increased dramatically over the recent years. The reasons of remaining single are various:some may be busy exploring careers without putting their marriage into the agenda; some may indulge in their jobs; travel; entertainment; physical fitness or friendship; More than 80% of them have not abandoned the value of marriage; and they say they aspire to marry or they want to be married someday; but they are patient and feel content being single until they meet the right person.Unit31. In China; home computer ownership is expected to exceed超过that of the U.S. and Europe combined. outstrip超过2. The company has been placed in a state of uncertainty处于不确定状态 as the board decides what to do next. limbo处于不确定状态3. He wrote quickly; but from time to time he stopped停下来 and look out of the window. ceased停止4. Thousands of refugees left the country following紧随 the outbreak of civil war. in the wake of 紧随5. An airline spokesman stated that the safety of passengers was absolutely of the most importance最重要的.paramount 最重要的6. The use of animals in scientific tests raises difficult moral 道德的 questions. ethical道德的7. TV coverage of the match had to be extended延长 when it went into over -time.prolonged延长8. He’s a good person to be with if ever you’re in a tough艰难的situation. thorny 艰难的9. He seemed to be a born天生的liar. congenital天生的10. The operation was rather painful; but I felt a lot better afterwards之后. subsequently之后11. In the Netherlands euthanasia安乐死has already been legalized.12. Some observers doubt whether the peace process can be salvaged 挽救.13. Lack of essential minerals can cause deformity畸形 in unborn children.14. All the cars are tested for defects不合格 before they leave the factory15. We help people with mental or physical handicaps残障to find work16. She had an lingering挥之不去的sense of guilt for some time after breaking off her relationship with Henry.17. The Government has to grapple with努力克服 the problem of unemployment.18. He was arrested for allegedly据说;据称 stabbing his former wife19. To her relief she was acquitted无罪开释 of all the charges laid against her.20. The frontiers尖端科学 of medical knowledge are being pushed farther outwards as time goes on.翻译：People who are energetic; happy; and relaxed are less likely to catch a cold than those who are depressed; nervous; or angry. When the brain is “happy”; it sends messages to our organs that help keep the body healthy and sound. Your chance of developing the common cold; pneumonia; or even cancer may very well be decreased by keeping your brain in a healthy state. In addition; happy and relaxed people are prone to better health practices than their negative and stressed counterparts. They are more likely to get plenty of sleepand to engaged in regular exercise; and have been shown to have lower levels of certain stress hormones.Unit 41.Westerners perceive shape and dimensions规模;大小------size2. see cross-cultural varieties品种；种类------types3. any intrusion侵入；闯入------penetration4. essence本质;实质------core5. the word means “in a house” or “from a chamber室内的；私人的”------room6. adaptations适应；改编------adjustments7.The amount of distinguishing有区别的------distinctive8. make careful distinctions区别 among different spatial elements ------differentiation9. secular and sacred神的；神的holy10. different perceptions of time may cause cultural conflicts冲突；矛盾------clashes11. Time is not a “mere connection”; as some English anthropologists人类学家but one of the most basic organizing systems of life; because all situational behavior has a temporal and a spatial dimension.12. Patterns模式are those implicit cultural rules by means of which sets are arranged so that they take on meaning.13. In fact; compared with more tangible有形的；切实的assaults on his character------nam ely Jones’ pending sexual—harassment lawsuit and the federal investigation into White—water ------words in a book could barely hurt him.14. A small but persistent固执的;坚持的… still don’t buy it.15. Mr. Smith; who was worried that the ban might infringe侵犯；违反—abiding gun owners ; had already voted against the bill. 16. After all; scientists noted; HIV is a retrovirus; a class of infectious agents known for their alarming ability to integrate 使…完整；使…成整体17. Suddenly; perfectly secular世俗的；长期的medieval—era arguments against the hubris of science.18. My study of space---- how people experience it and create a model of the spatial空间的；存在于空间的world in the central nervous system-----forced me to acknowledge the perceptual clichés of my culture.19. My ignorance of both hydroponics and of florist’s花店shops made me feel somewhat ease.Consequently I did not communicate in the manner I generally use when I am speaking on a familiar subject in a familiar setting.20. To understand an organization; you have to consider all of its components. Organizations arenetworks of related parts. Each element works together with the others to support efficient operations. The new MBA buzzword for it all is Organizational .Architecture建筑学；建筑风格.翻译：As one travels abroad and examines the ways in which space is handled; startling variations are discovered; differrences which we react to vigorously. Since none of us are taught to look at space as isolated from other associations; feelings cued by the handling of space are often attributed to something else. In growing up people learn literarily thousands of spatial cues; all of which have their own meanings in their own contexts.当人们到海外旅游时;如果留心观察外国人如何处理空间关系;就会发现许多令人惊讶的不同之处；而这些不同之处总让我们反应强烈..因为从来没有人教我们要把空间与其它联想区别开来;所以我们常认为空间处理引发的感觉是出于其它的原因..从小到大..人们要学会几千种空间暗示;所有这些暗示在不同的情境中都有不同的含义..Unit 51.rapt complete; engrossed 全神贯注2.pay for the use of the inn’s facilities resources; means 设施3.condone overlook; ignore 宽恕;赦免4.abominable contemptible 令人憎恶的5.assessments appraisals评估6.her friend’ s mother was livid furious 狂怒的7.gallantly bravely; politely 勇敢的8.jerk idiot; fool 性情古怪的人9.provocation irritation;annoyance 挑衅10.deliberate intentional; purposeful 深思熟虑的11.It is not our desire to be provocative挑衅的; but the status quo cannot continue. Neither can the present standoff; without the danger of a more serious confrontation that nobody wants.12.Like all derivatives; they function; essentially; as bets on the direction of particular markets.So coveted is such insurance that the total face amount -----or notional概念性的;抽象的“ value”-----of swaps and similar contracts has soared to an astronomical $11 trillion.13.Annoyed; shareholders have moved to curb抑制suchabuses-----just as they have moved to remove a number of CEOs in recent months.14.The day was star—crossed; Friday; October 13th; on the eve of the second anniversary of an devastating灾难性的market crash. 15. In an appalling令人震惊的demonstration of the mistreatment of recruits and their meager food rations; in March 1993 four sailors on Russky Island in the Far East died of malnutrition.16. Even more interesting; from a business perspective远景;观点;角度;is the so—called intranet----the collection of networks that connect computers within corporations---- that both Sun and Microsoft have targeted as a rich area for growth.17. She was at times less successful winning over her counterpart 副本;相似的人或物at the U.N.; who; like Napoleon’s army; travels on his stomach.18. Members of the NASA—led team arrived in Washington fully prepared to enter the fray紧张.19. Negotiation could sometimes defuse平息;缓解these situations and produce more acceptable consequences for both parties.20. It may not seem much consolation to point out that the teacher; too; becomes frustrated 挫败的;泄气的when his efforts appear to produce less than obvious result.翻译：Wealth exposed American manners to continuing international scrutiny because it underwrote an expansion of tourism. In the late nineteenth century the American abroad symbolized for some the dangders of sudden wealth and became an instrument to expose comparative social codes and standards of morality. Boasting; vulgarity; obsessive acquisitiveness; and insensitivity were soon associated with this early version of the Ugly American.Representing a small portion of the population; the international travelers seemed to stand for the whole country in the eyes of many foreigners; and it was the prodigality of their wealth; the apparently limitless wallets and purses; that seemed most dangerous. American money threatened to loot the Old World of its historic treasures; and aggressive celebrity hunters and relic gatherers to invade European privacy as well.财富使美国人的行为一直受到国际上的关注;因为它是旅游业的基础..19世纪末;有人认为;在海外的美国人是那些骤增财富的内在危险的象征;并使他们那并非完美无缺的社会准则和道德标准面临腐蚀的危险..自吹自擂、粗俗鲁莽、贪得无厌、麻木不仁很快成了早期的丑陋美国人的象征..周游列国的旅行者在全国人口中虽然只是一小部分;但在许多外国人眼里;他们好像代表了这个国家的所有人..他们挥霍无度;看起来有用之不尽的资产;这是最大的危险..美钞大有把旧世界历史上的奇珍异宝掠之一空之势..一群追逐名利和涉猎文物古玩的人无孔不入;极大地扰乱了欧洲人的私生活..Unit 61. Marsgall’s skills and her pleasant and sociable personality made her very effective in her public relations job. outgoing对人友好的;开朗的2.Mary was small; pretty; and very ambitions. petite女子娇小的;小个子的3. He was always strongly and clearly…emphatically着重地；强调地4. On the surface Dana was calm and self-assured;…self-possessed 镇静的；冷静的5. Many of its best-educated residents have been driven to neighboring countries in searching of better jobs. quest追求；寻找6.He was certainly the most attractive and pleasant lecturer there. personable风度好的；貌美的7. For a while she imagined that ….fntasized幻想；想像8. I like the idea of getting married but I don’t accept the traditional view of what marriage should be all about. buy into买进；入股9. They decided to postpone that she wedding until Pam’s mother was out of the hospital defer推迟；延期10. The government involved in a bitter dispute between the miners and their employers. caught up抓起;捉住11. Fathers are encouraged to be more involved有关的；卷入的with their families.12. Edinburgh sees itself as a university of the new millennium; committed坚定的；效忠的to research and teaching.13. I couldn’t figure out解决；算出；想出who the lady with the sunglasses was.14. Researchers are convinced确信的；深信的that…15. Spatial mobility tends to be highest among the most affluent 富裕的；丰富的groups on the one hand and the poorest on the other.16. She didn’t want to be tied down冷静by a full-time job.17. She loves her life and is immersed in the quest追求；寻找for knowledge for knowledge’s sake.18. It seems like everybody in America is looking for the end of the rainbow虹桥的尽头.19. Some claim that the herb has therapeutic治疗的value for treating pain.20. She has lots of good ideas; but she has formulating使公式化；系统地阐述difficulty.翻译：我赞成这样的理想：人应该各方面都很出色;既要有头脑;也要有健美的体格..我希望能得到某种幸福和快乐..美国是能得到这些东西的最好的地方;因为这里的人和我们的思想来自世界各地..至今到美国来寻梦的仍然大有人在..现在我明白了;决定美国梦能否实现的不是教育、不是机会、也不是艰苦的工作;而是权势和恐惧感..你在企业里爬到的地位越高;你失去的东西就越多..美国梦并没有破灭..今天在美国十分流行的想法是：千万不能失去梦想..I agree to this ideal: people should be good in all aspects; It is necessary to have not only the mind; but also magnificent physique. I hope to get some kind of happiness and joy. The United States is the best place to get these things; because the people here and their ideas come from all over the world. So far ; here are still many people to seek a dream.Now I know; It is not education; not opportunities; and not hard work; but power and fear that decide whether the American dream can come to be or not. The higher your position is in enterprise; the more you lose something. The American dream is not the end. The idea is very popular in the United States today: never lose one’s dream.Unit 71. The old lady complained continuously不停抱怨 about the way her granddaughters dressed .carped 吹毛求疵2. His painting gave a concrete form给一个…形式 to the spirit of the age.embodied呈现3. The poor couple suffered great pain经受巨大痛苦 every night over the decision to send their sons to school and keep their only daughter at home to help with farming work.agonized感到巨大痛苦4. He issued an open and direct 公开直接的denial 否定of the story that he was involved in bribery. outright 直率的;彻底的5. Bill looked hard凝视 at the computer screen; wondering what his programming mistake was. peered凝视6. Mary’s parents knew clearly that their daughter’s poor 可怜的scoreswould keep her out of any prestigious university. mediocre 中等的;普通的7. The company was flooded with被…淹了application letters after their advertisement for new employees appeared in local newspapers. inundated 淹没8. The manager stubbornly opposed坚决反对 the suggestion to reduce his staff despite his failing business. bucked 抗拒9. John Bull is always speaking sharply尖锐的地说to his workers for no apparent reason. snarling at咆哮10. The rain came gradually to a stop逐渐开始停止before we got started. petered out逐渐消失11. Gas lamps became obsolete淘汰 when electric lighting became popular.12. It is very hard for me not to wince畏缩when I see a nurse puttinga needle into my arm.13. No one would deny that the marathon is the most grueling折磨人的event in the sports meeting.14. They whizzed through the rehearsal so that there would be time for a short meeting.15. The film star made an pretentious自命不凡的;做作的 speech to journalists; stating that he only cared about art;not wealth. 16. The senator squelched镇压 the reporters who tried to interrupt him during his speech.17. It took a long time for him to mull思索over the whole thing before eventually making a decision.18. It was exciting to see such a movie for the first time; but we soon became jaded厌倦when our TV was flooded with programs of a similar kind.19. Jane hoped her new housemaid could be trusted; but she will had some misgivings疑虑.20. When the war broken out; a large number of refugees crossed the border; seeking sanctuary避难所in the neighboring country.21. Fame happened almost overnight for the 25-year-old actress; but she has been unfazed坦然的by all the recent media attention. 22. The man ’s posed假装 as a health-worker in order to get into the old lady’s house; and then stole her money.23. Years ago; people would have scoffed嘲笑at the notion that robots would operate on patients instead of surgeons.24. Mrs. Williams was upset when she heard that her son had misbehaved行为不端at school.翻译:Clinton and Bush are different greatly from each other. Clinton attends in person to everything; trival of critical; He can memorize data well just like a computer; He and his advisors never fail to take part in the interminable meeting of politics or economy. On the contrary; Bush dislike a meeting that lasts over 15 minutes; and even loathes being bothered by confusing figures. Clinton has cared very much about public approval presented by polls since his early presidency; where Bush; in a sharp contrast; proclaimed at the outset of his election that he would make all decisions on his own.Clinton is an avid reader; but his successor is not. Joyful; casual and elegant; the ex-president time and again drinks a little; but the new president has been a teetotal since he abstained from it 14 years ago.Bush Jr. was born of a privileged WASP family in Texas; while the former; from an ordinary family; has tempered himself and developed his talents in hardships. He works with his wisdom and eloquence; which his successor is short of.。
运用目的论对比分析爱伦坡《黑猫》的两篇中文译本
最新英语专业全英原创毕业论文，都是近期写作1 中国文化元素在功夫熊猫中的体现及其翻译2 中西方语言和文化间的相互影响3 功能翻译理论关照下的新闻英语标题翻译4 阶级矛盾导致的爱情悲剧―<<呼啸山庄>>解读5 从fans和PK看西方文化对汉语言文化的冲击与影响6 哥特式风格特征在《远大前程》中的诠释7 浅析英文爱情诗的特点与翻译方法8 从电影《刮痧》看中西方文化差异9 简•奥斯汀《诺桑觉寺》中人物对爱情和婚姻的不同态度10 论新闻英语中隐喻的运用及其翻译技巧11 文档所公布均英语专业全英原创毕业论文。
原创Q 799 75 79 3812 Eco-Critical Reading of The Call of the Wild13 《红字》的悲剧成因分析14 A Study of AP’s News Reports from the Angle of Cooperative Principle15 英文歌词翻译的原则和技巧16 The Exploration of Tragic Fate of Tess17 《喜福会》中的女性身份重建18 对《德伯家的苔丝》中纯洁一词不同理解的分析19 清教主义对美国文化的影响20 浅议提高英语阅读速度的方法21 英汉基本颜色词的文化内涵对比研究22 浅析小学英语课堂的有效提问23 运用目的论对比分析爱伦•坡《黑猫》的两篇中文译本24 Differences Between Chinese and Western Cultures in Gift-giving Customs25 从CIF与CIP的比较看CIP的优势26 提高英语听力的有效策略及教学启示27 浅谈商业广告的翻译28 双关语在英语广告中的应用与翻译研究29 从功能对等的角度浅析汉语成语中数字的翻译30 从功能对等理论看《生活大爆炸》中幽默字幕的汉译31 美国宗教文化及价值观在其外交政策中的体现32 《宠儿》的非线性叙事模式33 《收藏家》中空间与人物心理关系的解读34 (英语系经贸英语)浅谈成本领先策略对于企业发展的影响－－以格兰仕为例35 接受美学视角下莫言《红高粱家族》英译研究36 《外星人》电影海报的多模态话语解析37 国外品牌翻译及其接受度研究38 An Analysis of Symbolism in The Awakening39 小议非英语专业大学生英语口语能力的培养40 解析名词化与商务语篇的汉英翻译41 浅议女性哥特主义在《暮光之城》中的体现42 论《儿子与情人》中的恋母情结43 论地理位置对中美民族性格的影响44 《鲁宾逊漂流记》与《桃花源记》中乌托邦思想之比较45 观《麦田里的守望者》的教育失败46 违反合作原则所表达的会话含义—以《越狱》中Theodore Bagwell话语为例47 Pecola’s Blues－－A Reading of The Bluest Eye48 Confuci anism’s Influence on Transcendentalism:Reflection on Emerson’s and Thoreau’s Philosophy49 电影片名翻译商业化所引发的问题及应对策略50 英汉思维方式差异对英译汉结构处理的影响51 The Comparison of the Two Main Characters in Daniel Defoe’s Roxana and Emily Zola’s Nana52 还《失乐园》中撒旦的本来面目53 华兹华斯与拜伦自由意识的主要区别54 试析翻译中的“假朋友”55 涉外商务用餐中的非言语交际56 从《一间自己的房间》分析弗吉尼亚•伍尔夫的女性主义思想57 古诗词英译关于夸张的翻译策略研究58 唯美主义与奥斯卡•王尔德的《道林•格雷的画像》59 我国高中生英语学习动机研究60 论《简爱》中话语的人际意义61 On Freudian Interpretation of Dreams In Sense and Sensibility62 是受害者还是恶棍——重新解读夏洛克63 Women’s Image in Pygmalion64 浅析福斯特《霍华德庄园》中的人文主义65 禅宗思想在艾米莉迪金森诗歌中的体现66 论《一个温和的建议》中的黑色幽默67 简析诚信在对外贸易中的作用68 论《呼啸山庄》中希斯克利夫的心路历程69 从《在路上》看“垮掉的一代”70 The Application of Symbolism in The Great Gatsby71 从合作原则角度解读《成长的烦恼》中的言语幽默72 A Brief Analysis of Public Sign Translation73 A Comparative Study on Lin Daiyu and Xue Baochai74 论索尔•北娄《赫索格》中现代人的焦虑75 英语专业本科毕业论文摘要的体裁分析76 A Comparison of the English Color Terms77 翻译的对等性研究及其应用78 The Alternation of Language: A Study of Microblogging V ocabulary79 商务英语写作中的语用失误研究80 论《黑夜中的旅人》中主人公的信仰冲突与融合81 从功能翻译理论看《风声》的字幕翻译策略82 任务型教学在初中英语的实施情况研究83 通往真正的人性之路：《紫色》中西丽的转变84 论世纪年代以来美国文化冲击对中国青少年的影响及教育策略改革的应对措施85 从《大卫科波菲尔》中看狄更斯的道德观86 伏尼契小说《牛虻》中主人公性格分析87 从《胎记》看霍桑对科学的态度88 从《老人与海》看海明威的人生观89 跨文化交际背景下英语禁忌语探析90 探究瓦尔登湖的积极现实意义——倡导和谐生存发展模式91 体育新闻标题汉译中修辞的作用——以历届奥运会为例92 英汉习语的概念隐喻对比研究93 On Translation Strategies of Online Subtitle Translation Group94 关联理论在中餐菜单英译中的应用95 论中美家庭教育——以《喜福会》为例96 翻译中的文化差异97 浅析托妮•莫里森《恩惠》中的母爱98 《海的女儿》中安徒生的悲剧情结分析99 Transcendentalism in Thoreau's Walden100 中英身势语中的文化差异101 《中国日报》与《纽约时报》灾难新闻之比——以系统功能语法为视角102 商务英语信函中的语用失误分析103 年代美国梦在《了不起的盖茨比》中的折射104 《瓦尔登湖》生态批评视角分析105 中西文化差异分析—以国际商务谈判为视角106 象征主义手法在《白鲸》中的运用107 英汉死亡委婉语的文化差异及其分类对比108 惠特曼的人文主义思想对美国现代诗歌创作和中国诗歌创作的影响——以《自我之歌》为例109 凯特•肖邦小说《觉醒》中的超验主义思想分析110 从电影《阿凡达》透视美国文化111 《飘》中生态女性意识的研究112 从《道连格雷的画像》谈唯美主义艺术观113 文化差异对习语翻译的影响114 《老人与海》中的象征主义115 分析《贵妇画像》中伊莎贝尔的个性特点116 Hawthorne’s Feminism Consciousness in The Scarlet Letter117 从《纯真年代》的女性角色看旧纽约的女性地位118 欧•亨利作品中的人生的价值探索119 合作学习在初中英语写作教学中应用的可行性研究120 论跨文化商务中的非言语交际121 论考琳•麦卡洛《荆棘鸟》中的女性形象122 《了不起的盖茨比》中象征主义的研究123 论《老人与海》中人与自然的矛盾性与和谐性124 从功能对等理论看汉语文化负载词的英译125 英汉问候语对比研究126 对《一小时的故事》的批评分析127 The Glossology and Translation of Rhetorical Devices of Harry Potter128 中西方婚礼礼服颜色的对比研究129 论英汉基本颜色词的文化内涵差异130 试析《野性的呼唤》中的生态观131 A Comparison of the English Color Terms132 互联网媒体对汉语纯度影响的研究133 The Comparison of Diet Culture between China and America134 A Comparison of the English Color Terms135 从传统节日庆祝方式的角度比较中英文化差异136 通过象征主义、梭罗的自然思想和梭罗的个人主义析《瓦尔登湖》137 A Comparative Study on the Celebrations of Traditional Chinese and Western Festivals 138 英语学习能力与风格的性别差异研究139 身势语在演讲中的重要性和运用研究140 论《麦田里的守望者》的意义141 浅谈中西体态语的差异142 英文外贸合同中表时间介词的使用和翻译143 房地产广告的英译研究144 浅析卡夫卡小说中的荒诞意识145 不伦，还是不朽？－－从柏拉图的哲学理论视角解读《洛丽塔》146 从中英文动物词汇看中西方文化差异147 生活在夹缝中的畸形人——评《小镇畸人》148 中西校园流行语的文化对比149 A Comparison of the English Color Terms150 企业资料的翻译原则151 项目教学法在英语写作课中的应用152 中英广告中的双关语探析153 外来词的翻译方法初探154 废墟上成长起来的南方新女性155 从欧•亨利笔下的小人物探寻人生的真正价值156 《老友记》中言语幽默的修辞学思考157 对当今中国大学英语作为第二外语的教学方法的观察和思考158 浅论广告英语的修辞特色159 《永别了，武器》中的自然象征意义160 分析女性语言特点在英语委婉语中的体现——以《绝望主妇》为例161 浅析莫里森名作《最蓝的眼睛》中查理的性格特点162 《彼得潘》中的“成长”主题163 中西方商务礼仪的差异164 A Brief Study of Anti-female Bias in English165 语言经济学视角下的商务英语信函写作166 汉语成语英译过程中动物形象的转换167 从生态女性主义视角解读《喜福会》168 从自然主义角度解读《苔丝》的悲观主义169 《傲慢与偏见》中的婚姻观170 An Analysis of Gothic Features in Jane Eyre and Wuthering Heights171 析《道林格雷》中王尔德用来揭示生活与艺术冲突的方法172 研究网络语言的特点，影响以及其规范性173 从《丧钟为谁而鸣》看海明威死亡情节成因174 浅析库尔特•冯尼古特《猫的摇篮》中的黑色幽默175 从中美送礼习俗分析两国文化价值观的差异176 谈目的论视角下的商标翻译177 英国湖畔派诗歌的语言特点及其汉译178 从餐桌礼仪看中美饮食文化差异179 问题类型对TEM阅读成绩影响的实证研究180 Which Woman is More Popular in Modern Society：A Comparative Study of Tess and Jane 181 从文化差异的四维度读解中法葡萄酒文化182 浅析跨文化交际中的体态语183 《喧哗与骚动》中没落的悲剧184 合作原则视角下中美情景喜剧中言语幽默的对比研究185 《麦琪的礼物》和《傲慢与偏见》中的婚姻观对比研究186 美国电影中的中华文化运用得与失的研究187 论赫尔曼•梅尔维尔《白鲸》中的象征主义188 模糊限制语在求职中的应用研究189 A Comparison between Scarlett O’Hara and Jane Eyre from the Perspective of Feminism190 “同一性危机”——浅析汉娜的悲剧人生191 Analysis of Social Condition from American Gold Rush in The Call of the Wild192 A Research on V ocabulary Learning Strategies Employed by Non-English Majors in the CALL Environment193 浅谈英汉颜色词在使用中的文化差异194 目的论视角下新闻标题汉译英研究195 影响中学生英语学习的心理因素分析196 从《爱玛》看简•奥斯丁的爱情观197 浅析《麦田里的守望者》中的部分重要象征物198 The Linguistic Features of American Inaugural Address199 从生态批评角度看艾米丽狄金森的诗歌200 英语习语的文化内涵。
中英文对照适用社会学
专业英语Sociological Terminologies 中英文术语对照第一部分 Part OneI-欧洲古典社会学家Auguste Comte 奥古斯特孔德Karl Marx 卡尔马克思Herbert Spencer 赫伯特斯宾塞Vilfredo Pareto 维尔弗雷多帕累托Ferdinand Toennies 费迪南德滕尼斯Emile Durkheim 埃米尔涂尔干Georg Simmel 格奥尔格齐美尔Gaetano Mosca 加耶塔诺莫斯卡Max Weber 马克斯韦伯Leonard T. Hobhouse 莱奥纳多;T;霍布豪斯Robert Michels 罗伯特米歇尔斯II-北美古典社会学家William Graham Sumner 威廉姆格拉汉姆萨姆纳Lester Ward 莱斯特沃德Albion Small 阿比奥斯莫尔Franklin Giddings 弗兰克林吉丁斯Thorstein Veblen 索斯坦凡勃伦George Herbert Mead 乔治赫伯特米德W. I. Thomas W;I;托马斯Charles Horton Cooley 查尔斯霍顿库利Robert E. Park 罗伯特;E;帕克E. A. Ross E;A;罗斯III-现代早期Pitirim Sorokin 皮特里姆索罗金Elton Mayo 埃尔顿梅约Georg Lukacs 格奥尔格卢卡奇William F. Ogburn 威廉姆;F;奥格本Karl Mannheim 卡尔曼海姆Alfred Schutz 阿尔弗雷德舒茨Herbert Blumer 赫伯特布鲁默Paul Lazarsfeld 保罗拉扎斯菲尔德George Gallup 乔治盖洛普Tarcott Parsons 塔尔科特帕森斯George Homans 乔治霍曼斯IV-现代晚期David Riesman 大卫里斯曼Robert King Merton 罗伯特金默顿Barrington Moore 巴林顿摩尔Lewis Coser 刘易斯科塞Reinhard Bendix 莱因哈特本尼迪克斯C. W. Mills C;W;米尔斯Harold Garfinkel 哈罗德加芬克尔Peter Blau 彼特布劳Daniel Bell 丹尼尔贝尔Tom Bottomore 汤姆博托莫尔Alvin Gouldner 埃尔文古尔德纳Erving Goffman 欧文戈夫曼Seymour Martin Lipset 塞默尔马丁李普塞Richard Emerson 理查德爱默森Ralph Dahrendorf 拉尔夫达伦多夫Neil Smelser 尼尔斯梅尔赛V-当代社会学家Norbert Elias 诺伯特埃利亚斯S. N. Eisenstadt S;N;艾森斯塔德James Coleman 詹姆斯科尔曼Michel Foucault 米歇尔福柯Aaron Cicourel 阿隆西科利尔Peter Berger 彼特伯格Jurgen Habermas 尤尔根哈贝马斯Jean Baudrillard 让布西亚Pierre Bourdieu 皮埃尔布迪厄Niklas Luhmann 尼克拉斯卢曼Randall Collins 兰多柯林斯Jeffrey Alexander 杰弗瑞亚历山大Immanuel Walerstein 埃曼努尔沃勒斯坦Anthony Giddens 安东尼吉登斯E. P. Thompson E;P;汤普逊Charles Tilly 查尔斯蒂利Herbert Gans 赫伯特甘斯Howard S. Becker 霍华德;S;贝克尔Gary S. Becker 加里;S;贝克尔John Goldthorpe 约翰戈德索普Manuel Castells 曼努埃尔卡斯特尔斯Claude Fischer 克劳德费希尔Michael Burawoy 迈克尔波拉维Michael Hechter 迈克尔赫克特Theda Skocpol 西达斯考奇珀尔VI-其它重要相关社会科学家Claude Henri Saint-Simon 克劳德亨利圣西蒙Thomas Malthus 托马斯马尔萨斯Alexis de Tocqueville 阿列克西德托克维尔Charles Booth 查尔斯布思Frederic Le Play 弗里德里克勒普莱Sigmund Freud 西格蒙德弗洛伊德Ferdinand de Saussure 费迪南德德索绪尔Edmund Husserl 埃德蒙德胡塞尔Alfred Radcliffe-Brown 阿尔弗雷德拉德克利夫-布朗Joseph Schumpeter 约瑟夫熊彼特Bronislaw Malinowski 布拉尼斯洛马林诺夫斯基Marc Bloch 马克布洛赫Antonio Gramsci 安东尼葛兰西Jean Piaget 让皮亚杰Karl R. Popper 卡尔;R;波普B. F. Skinner B;F;斯金纳Fernand Braudel 费尔南德布劳代尔Claude Levi-Strauss 克劳德列维-斯特劳斯Abraham Maslow 亚伯拉罕马斯洛John K. Galbraith 约翰;K;加尔布雷斯Margaret Mead 玛格丽特米德Jean Paul Sartre 让保罗萨特Roland Barthes 罗兰巴特Walt Rostow 沃尔特罗斯托Louis Althusser 路易斯阿尔都塞Ernest Gellner 恩内斯特盖尔纳A. Noam Chomsky A;诺姆乔姆斯基第二部分 Part TwoAction Theory 行动理论Annale School 年鉴学派历史学Annee sociologique Durkheimian School 社会学年鉴涂尔干学派Applied Sociology 应用社会学Atomism 原子论Behaviorism 行为主义Chicago School 芝加哥学派Clinical Sociology 临床社会学Comparative Sociology 比较社会学Concentric Zone Theory 同心圆理论Conflict Theory 冲突理论Convergence Theory 趋同论Critical Theory and Frankfurt School 批判理论/法兰克福学派Cultural Relativism 文化相对主义Dependency Theory 依附理论Determinism and Its Variants 决定论及其变种Economic Determinism 经济决定论Technology Determinism 技术决定论Geographical Determinism 地理决定论Biological Determinism 生物决定论Dramaturgy/Dramaturgical Sociology 拟剧说/拟剧社会学Dualism 二元论Elite Theory 精英理论Empirical Sociology 经验社会学Empiricism 经验主义End of Ideology Theory 意识形态终结论Ethnomethodology 常人方法学Evolutionary Theory 进化论Exchange Theory 交换理论Falsificationism 证伪主义Feminism 女权主义Feminist Social Theory 女权主义社会理论Field Theory 场论Figurational Sociology 构型社会学Fordism 福特主义Formal Sociology 形式社会学Formal Theory 形式理论Functionalism 功能主义Game Theory 博弈论Grand Theory 宏大理论Grounded Theory 扎根理论Hermeneutics 解释学Historicism 历史主义Holism 整体论Human Relations School 人际关系学派Humanism 人本主义/人道主义Idealism 唯心主义Individualism and its variants 个人主义及其变种Epistemological Individualism 认识论个人主义Religious Individualism 宗教个人主义Methodological Individualism 方法论个人主义Interpretative Sociology 解释社会学Labeling Theory 标签理论Macro-Sociology 宏观社会学Marxist Sociology 马克思主义社会学Materialism 唯物主义Historical Materialism 历史唯物主义Meaningful Sociology 意义指向的社会学Meta-Theory 元理论Micro-Sociology 微观社会学Middle-Range Theory 中层理论New Left 新左派New Right 新右派Normative Theory 规范理论Objectivism 客观主义Operationalism 操作主义Phenomenological Sociology/Social Phenomenology 现象学社会学/社会现象学Political Economy 政治经济学Positivism 实证主义Post-Fordism 后福特主义Psychologism 心理主义Rational Choice Theory 理性选择理论Realism 唯实论/实在论Sociological Realism 社会学唯实论/实在论Reductionism 还原论Relativism 相对主义Role Theory 角色理论Social Contract Theory 社会契约论Structuralism 结构主义Structural-Functionalism 结构功能主义Structuration Theory 结构化理论Symbolic Interactionism 符号互动论Systems Theory 系统理论Trait Theory 人格理论Voluntarism 唯意愿论第三部分 Part ThreeAction Research 行动研究Ad hoc Hypothesis 事后假设Aggregate 聚合体Aggregate Data Analysis 集合数据分析Analogy 类比Analysis of Variance ANOVA 方差分析Analytical Induction 分析性归纳Archaeology of Knowledge 知识考古学Attitude Scale 态度量表Bar Chart 柱状图Behavior Therapy/Modification 行为疗法/行为矫正Bias 偏误Biased Sample 有偏样本Case Study 个案研究Case Work 个案工作Cause, Causality, and Causal Relationship 起因、因果性和因果关系Causal Explanation 因果解释Causal Modeling 因果建模Census 普查Chi-square Test 卡方检验Classification 分类Cluster Analysis 聚类分析Cluster Sample 整群抽样Coding 编码Cohort/Cohort Analysis 同期群科伙/同期群分析Community Study 社区研究Comparative History 比较史Comparative Method/Quasi-Experimental Method 比较法/准实验法Concomitant Variation 共变关系Construct 概念建构物Content Analysis 内容分析Control Group 控制组Conversational Analysis 会话分析/谈话分析Correlation/Correlation Coefficient 相关/相关系数Cross-Sectional Study 横断面研究/定点研究Cross Tabulation/Contingency Tables 交互分类表/列联表Counterfactual 反事实臆想Definition 定义Delphi Method 德尔斐法Dependent Variable/Independent Variable 自变量/因变量Descriptive Statistics 描述性统计Deviant Case Analysis 特异个案分析Discourse 话语Distribution 分布Elaboration 详析Emergent Properties 突生性状Emic/Etic 本位法/客位法Empathic Understanding 同情理解Epistemology 认识论Error 误差Estimation 估计值Ethnography 民族志Evaluation Research 评估研究Experimental Method 实验法Explanation 解释Causal Explanation 因果解释Deductive Explanation 演绎解释Probablistic Explanation 概率解释Meaningful or Purposive Explanations 目的性/意图性解释Functional Explanations 功能解释Evolutionary or Ecological Explanations 进化论或生态学解释Teleological Explanations 目的论解释Explanatory Mechanism 解释机制Exploratory Data Analysis 探索性数据分析Factor Analysis 因子分析Family Therapy 家庭疗法Fieldwork 田野调查Frequency Distribution 频数分布Group Therapy 团体疗法Guttman Scale 哥特曼量表Hawthorne Effect 霍桑效应Histogram 直方图Historiography 历史志Hypothesis 假设Hypothetico-Deductive Method 假设演绎法Ideal Type 理想型/理念型Idiographic/Nomothetic Approach 以个体独特经验为导向的研究/以普遍法则为导向的研究Index 指数Individual Level Data 个体层次数据Induction 归纳Interaction of Variables 变量间互动关系Interpretation/Verstehen 解释/理解Inter-Subjectivity 互主观性/主体间性Intervening Variable 中介变量Interview 访谈Interviewer Bias 访谈者引发的偏误Kurtosis 峰度Life History Method 生活史研究Likert Scale 李克特量表Log Linear Analysis 对数线性分析Longitudinal Study 追踪研究Market Research/Commercial Ethnography 市场研究/商业民族志Measurement Criteria and Levels 测量标准和测量层次Nominal-Level Measurement 定名测度/定类测度Ordinal-Level Measurement 定序测度Interval-Level Measurement 定距测度Ratio-Level Measurement 定比测度Measures of Central Tendency 集中趋势测量Mode 众数Median 中位数Mean 平均数Measures of Dispersion/Variability 离散趋势测量Range 数值变动范围Variance 方差Standard Deviation 标准差Standard Error 标准误Skew 偏态Kurtosis 峰度Model 模型Multidimensional Scaling 多维度空间建模Methodology 方法论Multivariate Analysis 多变量分析Factor Analysis 因子分析Principle Component Analysis 主成份分析Cluster Analysis 聚类分析Multiple Regression Analysis 多元回归分析Multivariate Analysis of Variance MANOVA 多变量方差分析Log-linear Analysis 对数线性分析Naturalistic Research Method 自然主义研究方法Non-Parametric/Parametric Statistics 非参数统计/参数统计Non-Response 无回答Normal Distribution 正态分布Null Hypothesis 虚无假设Objectivity 客观性Occupational Scale 职业量表Ontology 本体论Operationalization 操作化Opinion Poll 民意调查Oral History 口述史Panel Study 定组研究Paradigm 范式Participant Observation 参与式观察Path Analysis 路径分析Pie-Chart 饼状图Pilot Study 前导研究Population 总体Qualitative Analysis 定性研究Quantitative Analysis 定量研究Questionnaire 问卷Quota Sample 配额样本Regression 回归Reliability 信度Replication 副本Respondent/Informant 调查对象Response Rate 回收率Sample 样本Sampling 抽样Probability Sampling 概率抽样Representative Sampling 代表性抽样Sampling Error 抽样误差Sampling Frame 抽样框Scaling 量表化Sensitizing Concept 敏感化概念Significance Test 显着性检验Snowball Sampling 滚雪球抽样Social Facts 社会事实Social History 社会史Social Indicators 社会指标Social Statistics 社会统计Sociometry/Sociogram 社会测量学SPSS 社会科学统计软件包Spurious Correlation 伪相关Statistical Analysis 统计分析Statistical Inference 统计推论Stratified Sample 分层抽样Subjectivity 主观性Survey 问卷调查Systematic Sample 系统抽样Taxonomy 分类学Teleology/Tautology 目的论解释/循环论证/同义反复Thick Description 深描Time Budget Studies 时间预算分析Time-Geography 时间地理学Time Series 时间序列Typification 分类Typology 类型学Unobtrusive Measures 非干扰性测量Validity 效度Value-Free 价值无涉Value Neutrality 价值中立Value Relativity 价值相对性Value Relevance 价值相关性Variable 变量Variance 方差World View/World Vision 世界观。
急慢性溃疡性结肠炎小鼠模型的建立与比较研究
广东药科大学学报Journal of Guangdong Pharmaceutical University Jan, 2024, 40(1)急慢性溃疡性结肠炎小鼠模型的建立与比较研究刘闯1，李依林1，左泽平2，田颖颖1，赵新月1，吕英楠1，徐意2，张硕峰1，王志斌1，2（1.北京中医药大学中药学院，北京 102401； 2.北京同仁堂科技发展股份有限公司制药厂，北京 100071）摘要：目的考察葡聚糖硫酸钠（DSS）诱导小鼠慢性溃疡性结肠炎模型与噁唑酮（OXZ）诱导小鼠急性溃疡性结肠炎模型小鼠生理病理变化，为溃疡性结肠炎治疗药物开发研究中动物模型的选择提供依据。
方法通过连续7周交替饮用含1%DSS饮用水以及动物正常饮水诱导C57BL/6N小鼠慢性溃疡性结肠炎；OXZ腹部涂抹致敏后，用40%乙醇溶解灌肠诱导Balb/c小鼠急性溃疡性结肠炎。
模型建立后，通过疾病活动指数（DAI）评分、脏器指数、血液学检查、结肠组织HE染色、ELISA法检测TNF-α、INF-γ以及IL-10表达水平评价造模效果并对两种模型进行比较研究。
结果经DSS及OXZ诱导的小鼠结肠炎（UC）模型中，模型组小鼠均出现体质量下降、稀便以及便血等UC临床症状，DAI评分显著升高（P<0.01），与对照组相比，模型小鼠胸腺指数显著减低（P<0.01）、脾脏指数显著升高（P<0.01）、结肠长度明显缩短（P<0.01）。
血液学检查以及相关炎症因子检测结果均可见模型组小鼠发生炎症反应；组织病理学检查可见模型组小鼠结肠出现溃疡、炎性细胞浸润等病理特征。
结论两种方法均可以诱导溃疡性结肠炎，但小鼠的生理状态与结肠病理变化有所不同，在实际实验中应根据研究目的以及药物治疗特点选择合适的模型开展实验。
关键词：溃疡性结肠；动物模型；葡聚糖硫酸钠；噁唑酮中图分类号： R-332 文献标识码： A 文章编号： 2096-3653（2024）01-0095-06DOI: 10.16809/ki.2096-3653.2023101201Establishment and comparison of acute and chronic ulcerative colitis mice modelsLIU Chuang1, LI Yilin1, ZUO Zeping2, TIAN Yingying1, ZHAO Xinyue1, LYU Yingnan1, XU Yi2, ZHANG Shuofeng1*, WANG Zhibin1,2*(1.Chinese Medicine College of Beijing University of Chinese Medicine,Beijing 102401,China;2.Beijing Tongrentang Science and Technology Development Company Limited. Pharmaceutical Factory, Beijing 100071, China)*Corresponding author ZHANG Shuofeng, Email: shuofengzhang@; WANG Zhibin, Email: wangzhibin4804@ Abstract: Objective To investigate the physiological and pathological changes of mice models of chronic ulcerative colitis (UC) induced by DSS and acute ulcerative colitis induced by OXZ, and to provide evidence for the selection of animal model for the development of therapeutic drugs for ulcerative colitis. Methods Chronic ulcerative colitis was induced in C57BL/6N mice by alternating drinking water with or without 1% DSS for 7 consecutive weeks. Acute ulcerative colitis was induced in Balb/c mice by abdominal sensitization with OXZ combined with ethanol enema. After the establishment of the model, the modeling effect was evaluated through disease activity index (DAI) score, organ index, hematological examination, colon tissue HE staining, and ELISA detection of TNF-α, INF-γand IL-10 levels, and a comparative study was conducted between the two models.Results In the DSS and OXZ induced mouse colitis model, the model group mice showed clinical symptoms of UC such as decreased body weight, loose stools, and bloody stools. The DAI score was significantly increased (P<0.01). Compared with the control group, the thymus index of the model mice was significantly reduced, the spleenindex was significantly increased, and the colon length was significantly shortened (P<0.01). The results of收稿日期：2023-10-12基金项目：国家重点研发计划（2019YFC1711400）；国家科技重大专项（2018ZX09201-011）；国家科技重大专项（2018ZX09301-011-003）作者简介：刘闯（1999－），男，在读硕士生，研究方向为中药学，Email：******************通信作者：张硕峰（1969－），男，教授，主要从事中药药理研究，Email：**********************；王志斌（1966-），男，研究员，博士生导师，主要从事中药质量标准、中药药理方向研究，Email：**********************。
顺铂、奥沙利铂或奈达铂在Ⅲ期食管癌新辅助化疗疗效中的对比研究
[8]ZHANG L,BAO LY,TAN YJ,et a_Diaanostic peCormanco usingautomated breast ultrasonud system Hr breast cancer in Chinesewomeu ageU47yeac or older：A comparative study[J].Ultra-sonug MeU BHl,2019,45(12)：5157-3144.[19]ALMAHARI/MF,MAYWOOD MJ,LEVIEIO RB,et a_Maypingof metastatic level I axillap lymph dodos in patiexts with neclydiaanc/eU breast cancer-J].Int J Radiat Oncol Biol Phys,2620,106(4)：811-820.[20]乐坚，周世崇,范亦武，等•新辅助化疗后乳腺癌腋下淋巴结的超声影像分析[J].中华医学超声杂志(电子版),2610,13(7)：538-541.LE J,ZHOU SC,FAN YW,et al.Ultrasodograghic analysis of axillap lymph dodos in breast cancer after neoadjuvant cUemothecpy[J].Chinxe Joncal of MeUical Ultrasonud,2016,13(7)：538-541.[21]LI L,MORI S,KODAMA M,SAKAMOTO M,e(ad EndanceUsobograghic imagHg to diagncxo lymph node metastasis:Importance of blood vessel volume and densiqt J].Cancer Ros,2013,75：2982-2902.[22]BANYS-PALUCHOWSKI M,GRUBER IO,HARTKOPF A,eta_Axillap ultrasoupd Hr preUiction of wspcwse to deoadjuvanttheragy in Pc context of surgical strategics to axillap dissection inprimap breast cancer：A systemaXc review of the currext literature[J].Ach GyoecW Obstet,4929,361(2)：3/1-353.[23]BOUGHEY JC,SUMAN VJ,MITTENDORF EA,e(a_Sextineilymph godo surgep after deoadjuvant cUemothecpy in padeumwith godo-positive breast cancer：the ACOSOG Z1071(Alliance)clinical PH_J].JAMA,2013,310：3455-1461.[22]KUEHN T,BAUERFEIO JFEHM T,e(a_Sextinei lymph-nodebiopsy in patiexts with breast cancer before and aHer neoadjuvantcUemotherapy(SENTITA/:A prospective,multiceutro cohortstuPy-J].Lancet Oncol,2015,14(7):700-618.(编校：张西敏)顺钳、奥沙利钳或奈达钳在m期食管癌新辅助化疗疗效中的对比研究魏长江，茅怡铭，秦元，洪登森Comparative study oS neoadjuvaO cUemotaerapy for cisplaen,oxaliplahn or nebaplahn in stage m esophageal cancerWEI ChanOiang,MAO Yiming,QIN Ypgn,HONG DengsenThorncte Surgerg,Suzhou Kowloou Hospital Affiliated i Shanghai11(00隅UnfeTspi School cf M edicine,J f insp Suzhou015202,Chinn.[Abstract]Objective:To compare the efficacy and safety of heoadjuvant cUemotheragy Hr cisplatix,oxaliplatixand yeUap_Xx ix stage m esobdagevi cancer.Methoss:A pcspecOve enrollmenj of172paXents with heoadjuvantchemothempy Hr stage m esodhageal cahcer who met the inclusHu cPteCa duPng the pePod from Januap2214toJanuap2218ix Suzhou Kowloon Hospital AAXiateU to Syangdai Jiaotoug University School of MePiciho was per-HrmeU.All paXents undewvent surgep to analyze their ePicacy and dysfunctiox.The reaction occurreU.The neoadjuvant chemotheppy regimen was paclitaxel plus cisylaXh(n=53),cxaXymtih(n=37)or yeUapmXy(n=59),exe cycle everp3weePs,a total of9cycles of heoadjuvant cUemotheragy.Restlts：Among all paXents,patients receiveUCR,14)paXents with PR,59paXents with SD,and4paXents with PD.The objective ePective rate was63.5%.Therewas no significaht diRechce ix the overall survival betmeon the three groups(P二5462).Adverse cacOoxs ihcluPeUmyemsuppcssHx,_vef damage,mnal impairment,vomiting,and diarrbev.The1/01X6/00of grade m-V adverse c-actioxs was3.6%(14/172):CompaPsox between groups sPoweU that ueUaglaXh and oxxliplaXx were as ePective ascisplaXx,but the iycideyce of creaSniuc and vomiting was sig/ficanny lower than that of cisplatix,and the iycideyceof oxxXplanx diarrbev was siynifica_ny higher than the other two groups.Conclusion:NiUaglaXx and oxaliplaXh arenot infePor to cisplaXx ix heoadjuvant chemotheppy Hr stage m esobdagevi cancer,but the adverse reaction rate issiynHca_ny redpceU.【Key wods】esophaaegi cancer,heoadjuvant cUemotherany,cisplaXx,oxaliplaXh,ueUaglaXhModern Oncology222429(2))：0237-0262【收稿日期】2920-08-23【作者单位】上海交通大学医学院附属苏州九龙医院心胸外科，江苏苏州21500【作者简介】魏长江（1979-）,男见满江阴人，博士在读见住任医师,主要从事胸部肿瘤方面的临床研究工作。
中美初中数学教材几何习题的比较研究——以《勾股定理》为例
中美初中数学教材几何习题的比较研究——以《勾股定理》为例研究生姓名：李悦导师姓名：周莹教授学科专业：学科教学·数学研究方向：数学课程与教学论年级：2016级中文摘要随着课程和教育的改革，比较研究始终是教育教学研究的热点。
几何内容是数学教学的主要内容之一，各界对它的重视与日俱增。
教材习题作为教材内容的重要部分，引导学生理解知识内容、构建知识体系、深入理解知识、提高解题能力、发展数学思维、增强数学素养。
因此，国际间几何习题的比较研究对于借鉴别国优势有重要的意义和作用。
现有研究中，专门针对几何习题的比较研究寥寥无几，故有必要通过中外对比研究为我国初中数学教材几何习题的编写建言献策。
本研究选取美国CM版和中国人教版教材，以几何部分“勾股定理”一章的习题为例进行比较研究。
首先用文献分析法对“数学教材习题的相关研究”和“数学教材几何习题的比较研究”进行综述，了解已有研究，得到习题的表层研究趋于完善、几何习题深层比较维度有待挖掘、范希尔几何思维水平理论日益凸显等关键启示，为本文的比较研究奠定基础。
其次是中美两版数学教材的整体比较。
用内容分析法结合比较研究法从“整体结构”、“内容编排”和“所选章节内容设置”三个方面对中美两版教材进行整体比较分析，明确两版教材主体内容的异同，旨在为习题的比较研究作铺垫。
再次对中美两版初中数学教材几何习题进行比较研究。
从编排方式、数量、文字特征、插图、解题策略等维度对两版教材“勾股定理”习题进行表层比较分析，再基于范希尔几何思维水平理论，结合统计知识，从习题的背景、几何思维水平两个维度进行深层比较分析，得到两版教材几何习题的共同点：两版教材均分层设置习题；注重独立解题；习题的几何思维水平较高；但忽略了信息技术；缺乏合作意识；缺乏与其他学科的交互融合。
差异点：CM版重探索过程，人教版重解题能力；CM版注重强调关联知识内容；CM版更注重一题多用，注重还原现实生活中的原有情境；CM版插图类型更丰富，人教版较缺乏“实物”意识；人教版习题的几何思维水平更高。
恐龙研究方法英文作文
恐龙研究方法英文作文英文：When it comes to studying dinosaurs, there are avariety of methods that paleontologists use to gain abetter understanding of these prehistoric creatures. One of the most common methods is through the study of fossils. Fossils are the remains or traces of ancient organisms that have been preserved in rocks or sediments. By examining these fossils, paleontologists can learn about the anatomy, behavior, and habitat of dinosaurs.Another method of studying dinosaurs is through comparative anatomy. This involves comparing the anatomy of dinosaurs to that of living organisms to better understand their relationships and evolutionary history. For example, by comparing the bones of a T. rex to those of a modern bird, paleontologists can learn about the evolutionary link between the two groups.In addition to these methods, paleontologists also use computer modeling and simulations to study dinosaurs. By creating digital models of dinosaurs and their environments, scientists can simulate how these creatures moved, hunted, and interacted with their surroundings.Overall, the study of dinosaurs is a complex and multifaceted field that requires a combination of different methods and approaches. By using a variety of techniques, paleontologists can gain a more complete understanding of these fascinating creatures and the world they lived in.中文：当谈到研究恐龙时，古生物学家使用各种方法来更好地了解这些史前生物。
沟道宽度对65nm金属氧化物半导体器件负偏压温度不稳定性的影响研究
沟道宽度对65nm金属氧化物半导体器件负偏压温度不稳定性的影响研究崔江维;郑齐文;余德昭;周航;苏丹丹;马腾;魏莹;余学峰;郭旗【摘要】随着MOS器件尺寸缩小,可靠性效应成为限制器件寿命的突出问题.PMOS晶体管的负偏压温度不稳定性(NBTI)是其中关键问题之一.NBTI效应与器件几何机构密切相关.本文对不同宽长比的65 nm工艺PMOSFET晶体管开展了NBTI试验研究.获得了NBTI效应引起的参数退化与器件结构的依赖关系,试验结果表明65nm PMOSFET的NBTI损伤随沟道宽度减小而增大.通过缺陷电荷分析和仿真的方法,从NBTI缺陷产生来源和位置的角度,揭示了产生该结果的原因.指出浅槽隔离(STI)区域的电场和缺陷电荷是导致该现象的主要原因.研究结果为器件可靠性设计提供了参考.【期刊名称】《电子学报》【年(卷),期】2018(046)005【总页数】5页(P1128-1132)【关键词】65 nm;负偏压温度不稳定性;沟道宽度【作者】崔江维;郑齐文;余德昭;周航;苏丹丹;马腾;魏莹;余学峰;郭旗【作者单位】中国科学院特殊环境功能材料与器件重点实验室,新疆电子信息材料与器件重点实验室,中国科学院新疆理化技术研究所,新疆乌鲁木齐830011;中国科学院特殊环境功能材料与器件重点实验室,新疆电子信息材料与器件重点实验室,中国科学院新疆理化技术研究所,新疆乌鲁木齐830011;中国科学院大学,北京100049;中国科学院特殊环境功能材料与器件重点实验室,新疆电子信息材料与器件重点实验室,中国科学院新疆理化技术研究所,新疆乌鲁木齐830011;中国科学院大学,北京100049;中国科学院特殊环境功能材料与器件重点实验室,新疆电子信息材料与器件重点实验室,中国科学院新疆理化技术研究所,新疆乌鲁木齐830011;中国科学院大学,北京100049;中国科学院特殊环境功能材料与器件重点实验室,新疆电子信息材料与器件重点实验室,中国科学院新疆理化技术研究所,新疆乌鲁木齐830011;中国科学院大学,北京100049;中国科学院特殊环境功能材料与器件重点实验室,新疆电子信息材料与器件重点实验室,中国科学院新疆理化技术研究所,新疆乌鲁木齐830011;中国科学院特殊环境功能材料与器件重点实验室,新疆电子信息材料与器件重点实验室,中国科学院新疆理化技术研究所,新疆乌鲁木齐830011;中国科学院特殊环境功能材料与器件重点实验室,新疆电子信息材料与器件重点实验室,中国科学院新疆理化技术研究所,新疆乌鲁木齐830011【正文语种】中文【中图分类】TN386.11 引言负偏压温度不稳定性(Negative Bias Temperature Instability,NBTI)是纳米MOS 器件和集成电路最主要的可靠性问题之一[1,2].器件上施加的电压应力在较长时间或者高温加速作用下,会产生陷阱电荷,使器件的阈值电压漂移,跨导降低,亚阈摆幅增加,并且退化幅度与应力时间服从指数关系.这些参数退化可能会导致模拟电路高精度晶体管对的失配,从而影响电路的性能要求;阈值的漂移影响数字逻辑的实现甚至产生逻辑错误;漏电流下降影响驱动电路的驱动能力;跨导的退化同样影响响应速度或者引起时序问题.国内外非常重视纳米器件的NBTI效应研究,开展了较多的研究工作[3～9].已获得NBTI造成器件参数退化的现象,但对其缺陷形成机制及详细过程依然存在争议.NBTI效应有两种基本的退化模型:(1)界面态相关的模型,如最被广泛接受的反应扩散(Reaction-Diffusion,RD)模型,认为硅-二氧化硅(Si-SiO2)界面的Si-H键在应力条件下被打破,生成的H(氢)物质向氧化层中输运[3];(2)空穴俘获模型(hole-traping model),包括导带价带载流子发生弹性遂穿等,认为空穴通过遂穿进入氧化层,被陷阱俘获和释放,使得器件发生退化和恢复[4].在NBTI效应引起的参数退化中,阈值电压VTH漂移是最受关注的焦点,有研究报道了其退化幅度与沟道长度的关系[10～14].结果显示，NBTI效应引起的阈值电压变化会随着沟道长度L的减小而增加.研究认为产生该现象的主要原因是,与长沟道器件相比,短沟器件的栅氧化层边缘所占的比重更大,而这一区域往往具有更高的缺陷密度.另外,为了降低热载流子效应而采取的轻掺杂漏(Lightly Doped Drain,LDD)结构,使得栅氧化层边缘处具有更高的空穴浓度,也使得短沟器件更容易受到影响[12].与之对应的是,除少量文献报道了深亚微米器件NBTI损伤与沟道宽度的关系[13,14],很少有报道纳米器件NBTI效应与沟道宽度W是否存在相关性.基于以上背景,本文以65nm工艺的PMOSFET为研究对象,重点关注了NBTI效应在沟道宽度方向上的变化.2 试验条件试验样品为65nm工艺的体硅PMOSFET(P沟道金属氧化物半导体场效应晶体管).样品的栅氧厚度为1.8nm,正常工作电压VDD为1.2V.试验样品为封装器件,一只封装的器件中包含有4种沟道宽长比(W/L)的晶体管,其W/L分别为10μm /0.06μm、0.3μm/0.06μm、10μm/0.3μm、0.6μm/0.3μm.NBTI试验温度为400K,栅极电压VStress=-2V,其余端接地,应力时间为1000s.参数测试时,由于NBTI具有快速恢复的特性,中断应力测试会使得部分退化恢复.因此本试验中采用文献[15]中的方法,只测量一个点VGM处的电流IDLIN,通过IDLIN的电流变化计算阈值电压漂移:(1)其中,VTH0是在应力前扫描转移特性曲线,通过最大跨导处线性外推得到的阈值电压,IDLIN0是测量电压VGM处的初始电流.同时，在应力前后扫描转移曲线.为了尽量减小由于温度导致的误差,首先在升温的过程中多次扫描转移特性曲线,找到样品的零温度系数点,并将应力间的测量电压VGM设在零温度系数点附近,通过这种方法来避免温度引起的电流变化,保证测试数据的精确性.在多次测试时,将每次的升温、降温时间都保持一致,以此来保证多次试验的一致性.试验期间的电参数测试以及NBTI应力在Keithley 4200-SCS半导体参数分析仪上完成.在正式试验前,首先对同一批次的1只器件进行摸底试验,用来确定应力电压、测量电压、升温时间、降温时间等试验条件.正式试验时,选取同一批次的3只器件分别开展试验.试验样品和测试系统如图1所示.其中,图1(a)为被测器件,采用DIP48封装,内含不同宽长比的PMOSFET.图1(b)为测试系统,图中左侧为添加高温应力的试验箱,右侧为提供NBT应力和提供测试的半导体参数分析仪Keithley 4200-SCS.试验开始时,将器件放入高温箱,使用连接线将器件与半导体参数分析仪相连,进行应力试验和测试.3 试验结果对试验样品进行NBTI试验前后均在室温下扫描了IDS-VGS曲线,如图2(a)的W/L=0.6μm/0.06μm器件试验结果所示,器件的线性电流IDLIN下降,阈值电压VTH负向漂移.图2(b)是不同宽长比器件的阈值电压在应力期间随时间的变化,基本满足对数关系.表1为试验样品在NBT应力后的直流参数变化百分比.首先,用应力试验后的参数值减去参数初值,得到参数变化值.之后用参数变化值除以参数初值,得到参数变化百分比.对比不同宽长比的样品,可以看出窄沟样品几乎所有参数的退化都比宽沟道样品更大.这一现象与国外已报道的微米级PMOSFET的表现截然相反[13].3次试验中的3只器件均表现出一致的参数变化趋势.表1 65nm样品1000s应力后室温下直流参数变化百分比W/L (μm/μm)10/0.060.6/0.0610/0.30.6/0.3阈值VTH2.26%5.68%2.30%3.38%线性电流IDLIN4.07%6.37%3.39%4.39%饱和电流IDSAT4.54%6.61%5.15%6.53%最大跨导Gmmax2.83%2.27%2.46%2.05%4 讨论4.1 缺陷作用分析缺陷的存在是纳米器件可靠性效应引起参数退化的原因.NBTI效应引起器件参数退化的主要原因是应力导致了缺陷的产生,从而使器件参数随时间呈现出不稳定性.根据已有研究,NBTI退化主要源于界面处Si-H键发生断裂而产生界面陷阱的过程,此外栅氧化层陷阱的空穴俘获机制也是NBTI效应的另一个重要来源.在以SiO2或高K材料作为栅介质层的技术中,氧化层衬底界面存在着Si-O键和Si-H键.由于Si-O键具有较高的激活能因而需要与携带高能的载流子反应才可发生断裂,而Si-H键只要在靠近界面的冷空穴的辅助下就可以发生断裂反应.Si-H键在高温和电场应力条件下就可以发生断裂,从而形成界面陷阱[3],进而使器件参数发生退化.R-D模型认为在NBTI退化过程中Si-SiO2界面处发生了电化学反应,使得原本不具备电学活性的钝化Si-H键发生断裂生成界面陷阱和可自由移动的氢物质: Si-H+h+↔Si++Xinterface(2)可能的反应包括空穴辅助断裂和掺杂激活断裂[16,17]:Si3≡Si-H+h+↔Si3≡Si*+H(3)Si3≡Si-H+H+↔Si3≡Si*+H(4)另一方面,在靠近界面的氧化层中存在着氧化物陷阱E’中心,可以克服势垒形成氧空位结构,势垒的高度以及E’中心的稳定性则与其所处的网状结构相关.E’中心在氧化层的热生长过程中形成,其能级位于禁带的中部,因而它们的作用类似于空穴陷阱,可以俘获从反型层而来的空穴.对于NBTI效应来说,氧化物陷阱曾被认为是阈值电压不稳定的根源,随着MOSFET中氮元素的引入,该观点也被更多的地用以解释NBTI应力下阈值电压的退化.部分E’中心可以与Si-SiO2界面反复交换电荷,引起参数的退化与愈合.温度和电场是NBTI效应中缺陷产生的重要条件.在固定温度下,电场对Si-H键的断裂和氧化物中电荷的捕获非常重要,高电场更易于产生陷阱电荷.下文将通过电场产生位置及电场大小来分析NBTI的窄沟增强效应.4.2 NBTI退化随沟道宽度变化的原因在纳米MOS器件中存在两个主要的氧化层,即栅氧化层和浅槽隔离(Shallow Trench Isolation,STI)氧化层.在对试验样品施加NBT应力时,不仅栅氧化物中会产生电场,STI也会受到电场的影响.由于试验时栅端施加了高于正常工作电压的应力,在该应力条件下,栅氧化层下的沟道远比普通强反型状态聚集了更多的空穴.这些空穴的存在屏蔽了来自沟道上方的部分电场影响.而STI上方电场无法屏蔽,使得STI 与反型层界面也存在一定应力.NBTI效应作用位置,除了垂直电场影响的栅氧化层,沟道指向STI界面方向也存在电场影响STI区域.图3(a)为TCAD仿真工具构建的器件结构.图3(b)模拟了沟道指向STI方向的电场,可以看出水平方向电场在很大的STI界面区域上都存在,而且越靠近栅氧的区域电场越强.尽管与栅氧相比这一区域的电场强度较小,但是STI与栅氧不同,氧化层质量较差,缺陷密度也更高,甚至有研究报道STI产生的机械应力会导致STI边缘处产生更多的界面陷阱.尤其需要关注的是在LDD与STI接触的四个边角,因为其空穴浓度比沟道中心更大,四个边角位置的NBTI损伤都更大.与之相反的是,由于STI的存在,由沟道指向栅氧层的电场,在靠近STI区域比中间位置更小,也有报道认为栅氧靠近STI位置的厚度增加,是造成这一现象的原因[17].但是在我们的仿真中,栅氧的厚度是均匀的,仍然有同样的现象,如图3(c)所示.这或许可以解释文献[8]中与我们实验所获得的不同结果.文献[8]中的器件的栅氧厚度为3.5nm,沟道长度为0.18μm,沟道宽度分别为30μm、10μm、5μm、1μm.器件栅面积足够大,W的减小不足以使四个边角的比重获得大幅提高,反而使得栅氧上弱电场区域比重增加,此时NBTI的退化变小;本文中65nm器件的栅氧厚度为2.6nm,最小沟道长度仅为0.06μm,沟道宽度分别为10μm、0.6μm、0.3μm.器件本身栅氧面积较小,边角的比重随着W的减小迅速增加,其作用超过了栅氧上弱电场区域,所以退化更大,从而使得在相同NBT应力下,相同沟道长度的窄沟器件产生的参数退化更大.5 结论随着MOS器件特征尺寸缩小至纳米级,NBTI成为影响器件可靠性的重要问题之一.为了研究器件结构对NBTI退化的影响,本文开展了不同宽长比的65nm工艺PMOSFET的NBTI效应研究.试验结果表明,窄沟器件几乎所有的NBTI参数退化都要大于宽沟器件,这与国外已报道的微米级器件的现象相反.通过缺陷分析和TCAD 仿真的方法,分析了产生这一现象的原因.认为STI区域对器件NBTI效应有重要影响.由于纳米器件栅氧面积小,边角的比重随着沟道宽度的减小迅速增加,其作用超过了栅氧上的弱电场区域,因而引起更大的退化.研究结果为器件可靠性设计提供一定的参考.参考文献【相关文献】[1]Mahapatra S,Alam M A,Bharath K P,et al.Negative bias temperature instability in CMOS device[J].Microelectronic Engineering,2005,80(17):114-121.[2]郝跃，刘红侠.微纳米MOS器件可靠性与失效机理[M].北京:科学出版社,2008.1-18.[3]Huard V,Denais M，Parthasarathy C.NBTI degradation:From physical mechanisms to modeling[J].Microelectronics Reliability,2006,46:1-23.[4]S Mahapatra,N Goel,S Desai,et al.A comparative study of different physics-based NBTI models[J].IEEE Transactions Electron Devoces,2013,60(3):901-916.[5]张月.PMOSFET器件NBTI效应的物理机制与模型研究[D].西安：西安电子科技大学，2014. Zhang Y.A study on NBTI effect mechanism and modeling with PMOSFETdevices[D].XiAn:Xidian University,2014.(in Chinese)[6]Cao Yan-rong,Yang Yi,Cao Cheng,He Wen-long,et al.Recovery of PMOSFET NBTI under different conditions[J].Chinese.Physics.B,2015,24(9):097304-1-5.[7]张城绪.小尺寸pMOSFET器件的NBTI寿命预测方法研究[D].南京：南京大学，2016.Zhang X C.Study on NBTI Lifetime Prediction Method of Deeply Scaled pMOSFET Devices[M].Nanjing:Nanjing University,2016.(in Chinese)[8]Subrat Mishra,Hiu Yung Wong,Ravi Tiwari,et al.Predictive TCAD for NBTI stress-recovery in various device architectures and channel materials[A].2017 IEEE International Reliability Physics Symposium (IRPS)[C].Monterey,CA,USA:IEEE press,2017.6A-3.1-6A-3.8.[9]Theano A.Karatsori,Christoforos G,Theodorou.Characterization and modeling of NBTI in nanoscale ultra thin body ultrathin box FD-SOI MOSFETs[J].IEEE Transactions on Electron Devices,2016,63(12):4913-4918.[10]曹建民,贺威,黄思文,张旭琳.pMOS器件直流应力负偏置温度不稳定性效应随器件基本参数变化的分析[J].物理学报,2012,61(21):217305-1-8.Cao J M,He W,Huang S W,Zhang X L.Dependence of the DC stress negative bias temperature instability effect on basic device parameters in pMOSFET[J].Acta Physica Sinica,2012,61(21):217305-1-8.(in Chinese)[11]Toyoji Yamamoto,Ken’ichi Uwasawa,tohru mogami.Bias temperature instability inscaled p+polysilicon gate p-MOSFET’s[J].IEEE Traransactions ElectronDevices,1999,46(5):921-926.[12]Lei Jin,Mingzhen Xu.Effect of channel length on NBTI in sub-100nm CMOS technology[A].2008 2nd IEEE International NanoelectronicsConference[C].Shanghai,China:IEEE Conference Publications,2008.597-600.[13]G Cellere,M G Valentini,A Paccagnella.Effect of channel width,length,and latent damage on NBTI[A].2004 IEEE International Conference on Integrated Circuit Design and Technology[C].Austin,TX,USA:IEEE,2004.303-306.[14]CAO Yan-Rong,MA Xiao-Hua,HAO Yue,et al.Effect of channel length and width on NBTI in ultra deep sub-micron PMOSFETs[J].Chinese Physics Letters,2010,27(3):037301-1-4.[15]Schroder D K,Babcock J A.Negative bias temperature instability:road to cross in deep submicron silicon semiconductor manufacturing[J].Journal of AppliedPhysics,2003,94(1):1-18.[16]S Chakravarthi,A Krishnan,V Reddy,et al.A comprehensive framework for predictive modeling of negative bias temperature instability[A].In International Reliability Physics Symposium (IRPS)[C].Phoenix,AZ,USA:Reliability Physics SymposiumProceedings,2004.42nd Annual.2004 IEEE International,2004.273-282.[17]S N Rashkeev,D M Fleetwood,R D Schrimpf,et al.Defect generation by hydrogen at the Si-SiO2 interface[J].Physical Review Letters.2002,87(16):165506.。
DNA别构效应介导的蛋白质间长程相互作用的研究的开题报告
DNA别构效应介导的蛋白质间长程相互作用的研究的开题报告题目：DNA别构效应介导的蛋白质间长程相互作用的研究背景：DNA别构效应指的是DNA分子在不同条件下发生结构变化，如超螺旋、三维结构折叠等现象。
这些变化可能导致DNA与蛋白质结合时结合模式的改变，从而对细胞功能的调控产生影响。
DNA别构效应在基因表达、DNA修复等生命过程中起到重要作用。
近年来，关于DNA别构效应的研究引起了广泛关注。
研究目的：本研究旨在探究DNA别构效应对蛋白质间长程相互作用的影响机制，为阐明DNA-蛋白质相互作用的本质提供新的思路和实验数据。
研究方案：我们计划使用多种方法，包括NMR、X射线晶体学等，结合分子动力学模拟，探究DNA别构效应如何影响距离较远的蛋白质之间的相互作用。
具体的实验方案和流程如下：1. 选择两个蛋白质作为模型系统，其中一个具有DNA结合能力，另一个不具有DNA结合能力。
2. 借助NMR测定两个蛋白质在不同条件下的结构，并用X射线晶体学结合分子动力学模拟研究其相互作用方式。
3. 对比两种蛋白质在存在DNA时和不存在DNA时的相互作用，发现其是否发生变化。
4. 多种分析方法综合研究DNA别构效应对蛋白质间相互作用的影响机制。
5. 最终得到实验结果，对比分析，撰写成果报告。
预期成果：本研究的预期成果包括：1. 揭示DNA别构效应对蛋白质间长程相互作用的影响机制。
2. 拓展对DNA-蛋白质相互作用的认识，为探究生命过程的本质提供新的思路和实验数据。
3. 为调控基因表达、DNA修复等生命过程提供理论基础。
参考文献：1. Shokri L, Marcasite H, Cousin SF, et al. DNA breathing dynamics in the presence of (hyper) twisted proteins[J]. Nucleic acids research, 2019, 47(8): 3954-3962.2. Kusters I, Sanner M F J, de Leeuw R A W, et al. Understanding protein-DNA binding: Distinct structures with different energies[J]. Journal of chemical information and modeling, 2020, 60(4): 2179-2188.3. Gílbert R E, Rueda M, González C. A comparative study of two crystal structures of Lac repressor headpiece: Differences in hydration and DNA-binding reveal a breathing effect coupled to protein dynamics[J]. Proteins: Structure, Function, and Bioinformatics, 2018, 86(1): 45-56.。
- 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)。
A comparative study of two modeling approaches in neural networksZong-Ben Xu a ,Hong Qiao b ,Jigen Peng a ,Bo Zhang c,*aInstitute for Information and System Sciences,Xi’an Jiaotong University,Xi’an,China bDepartment of Computation,UMIST,P.O.Box 88,Sackville Street,Manchester M601QD,UK cSchool of Mathematical and Information Sciences,Coventry University,Coventry CV15FB,UKReceived 16April 2002;accepted 28May 2003AbstractThe neuron state modeling and the local field modeling provides two fundamental modeling approaches to neural network research,based on which a neural network system can be called either as a static neural network model or as a local field neural network model.These two models are theoretically compared in terms of their trajectory transformation property,equilibrium correspondence property,nontrivial attractive manifold property,global convergence as well as stability in many different senses.The comparison reveals an important stability invariance property of the two models in the sense that the stability (in any sense)of the static model is equivalent to that of a subsystem deduced from the local field model when restricted to a specific manifold.Such stability invariance property lays a sound theoretical foundation of validity of a useful,cross-fertilization type stability analysis methodology for various neural network models.q 2003Elsevier Ltd.All rights reserved.Keywords:Static neural network modeling;Local field neural network modeling;Recurrent neural networks;Stability analysis;Asymptotic stability;Exponential stability;Global convergence;Globally attractive1.IntroductionIn the current neural network research,two funda-mental modeling approaches are commonly adopted:either using the neuron states (the first approach)or using the local field states of neurons (the second approach)as basic variables to describe the dynamical evolution rule of a neural network.The recurrent back-propagation networks (Almeida,1988;Pineda,1987;Rohwer &Forrest,1987),for instance,provide a typical example of the first approach.The networks,as in their standard form and as a direct generalization of the well-known back-propagation network (Hertz,Krogh,&Palmer,1994),are modeled byt d vi d t ¼2v i þg i X N j ¼1w ij v j þu i 0@1A ;i ¼1;2;…;N :ð1ÞHere v i is the state of neuron i with u i ¼X N j ¼1w ij v j þu ibeing its local field,g i the activation function of neuron i ;u i the external input imposed on neuron i ;w ij the synaptic connectivity value between neuron i and neuron j ;and N the number of neurons in the networks.On the other hand,the famous Hopfield networks (Hopfield,1982;Hopfield &Tank,1986)are examples of the second approach and can be described in terms of the local field state u i ;i ¼1;2;…;N ;of neurons asC i d u i d t ¼2u i R i þX Nj ¼1w ij g j ðu j ÞþI i ;i ¼1;2;…;N ;ð2Þwhere v i ¼g i ðu i Þgives the state of neuron i ;and C i ;R i and I i are fixed physical parameters.We henceforth call the first approach the static neural network modeling and the second approach the local field neural network modeling .Correspondingly,a neural network modeled in terms of their neuron states such as Eq.(1)will be referred to as a static neural network model ,whilst a network modeled in0893-6080/$-see front matter q 2003Elsevier Ltd.All rights reserved.doi:10.1016/S0893-6080(03)00192-8Neural Networks 17(2004)73–85/locate/neunet*Corresponding author.Tel.:þ44-24-7688-8587;fax:þ44-24-7688-8080.E-mail addresses:b.zhang@ (B.Zhang);zbxu@ (Z.-B.Xu);hqiao@ (H.Qiao);jgpeng@ (J.Peng).terms of localfield states such as Eq.(2)will be referred to as a localfield neural network model.It is obvious that,given a neural network system,the two modeling approaches may both be applied to describe the system either from an external state point of view or from an internal state point of view.The questions then arise:would there exist any difference between these two modeling approaches?Which model would be more beneficial when a specific analysis purpose is concerned?Can these two different modeling approaches be cross-fertilized and complemented mutually(and if so how)?Answering such questions is clearly of great importance both theoretically and practically but,to the best of our knowledge,has not yet been given.The purpose of this paper is to answer some of these questions by establishing a comparison theory on dynamics of the two related models.Precisely,we will clarify the relationship between the dynamics of the static neural network model and the localfield neural network model in terms of their trajectory transformation property, equilibrium state correspondence property,global conver-gence and stability in many different senses.The results obtained will not only uncover the consistency and inconsistency of the two modeling approaches,but also lay a sound theoretical foundation for validity of a useful, cross-fertilization type approach to neural network analysis developed recently by the authors(Qiao,Peng,Xu,& Zhang,2003).In this paper we consider the following generic static neural network modelt d xd t¼2xþGðWxþqÞ;xð0Þ¼x0[R Nð3Þand the localfield neural network modelt d yd t¼2yþWGðyÞþq;yð0Þ¼y0[R Nð4Þwhere x¼ðx1;x;…;x NÞis the neuron states,y¼ðy1;y2;…;y NÞis the localfields,W¼ðw ijÞN£N is the synaptic weight matrix and G:R N!V#R N is the nonlinear activation mapping with V being a convex subset of R N:Depending on the specific application,the nonlinear mapping G may be defined component wisely or otherwise. In the former case,G is of the form:GðyÞ¼ðg1ðy1Þ;g2ðy2Þ;…;g Nðy NÞÞT;y¼ðy1;y2;…;y NÞT[R Nwith g i representing the activation function acted on neuron i;where A T stands for the transpose of A:It should be observed that the model(3)includes not only the recurrent back-propagation networks(Almeida,1988; Pineda,1987;Rohwer&Forrest,1987),but also other extensively studied neural networks such as the optimiz-ation type networks proposed by Bouzerdoum and Pattison (1993),Forti and Tesi(1995),Friesz,Bernstein,Mehta, Tobin,and Ganjlizadeh(1994),Liang and Wang(2000a,b),Xia(1996),Xia and Wang(1998)and Xia and Wang (2000),the brain-state-in-a-box(BSB)type networks(Li, Michel,&Porod,1989;Varga,Elek,&Zak,1996). Similarly,the Eq.(4)models not only the Hopfield-type networks(Hopfield,1982;Hopfield&Tank,1986),but also the bidirectional associative memory(BAM)type networks (Kosko,1988)as well as the cellular neural networks (CNNs)in Chua and Yang(1988),Park,Kim,Park,and Lee (2001)and Roska and Vandewalle(1995).It should also be noticed that the static neural network model(3)and the localfield model(4)can be correlated in a straightforward way if the matrix W is invertible(i.e. nonsingular).In this case,the two models can be transferred equivalently from one to the other.This trivial situation will not be considered here,so the nonsingularity of W will not be assumed in this paper.2.Relationship between dynamics of the two modelsIn this section the relationships between the dynamics of the models(3)and(4)are systematically clarified in terms of their trajectory transformation property,equilibrium correspondence property,stability,superior nontrivial attractive manifold property,asymptotic stability and global convergence.From the relationships clarified,a stability-invariance property of the models will be concluded in a sense to be specified later.We also point out some open questions related to the equivalence of the stability between Eqs.(3)and(4).2.1.Coherent systemsBesides the basic systems(3)and(4),we will also study the following four closely related systems:td x¼2xþGðWxþqÞ;xð0Þ¼x0[Co RðGÞð3:1Þtd xd t¼2xþGðWxþqÞ;xð0Þ¼x0[Co RðGW qÞð3:2Þtd yd t¼2yþWGðyÞþq;yð0Þ¼y0[RðW qÞð4:1Þtd y¼2yþWGðyÞþq;yð0Þ¼y0[RðWGÞþqð4:2Þwhere W q is the affine transformation defined by W q x¼Wxþq:These four systems are exactly the Eqs.(3)and(4) when restricted to the specific manifolds Co RðGÞ; Co RðGW qÞ;RðW qÞandþq;respectively,where RðGÞdenotes the range of operator G and Co RðGÞdenotes the closed,convex hull of RðGÞ(i.e.the smallest closed convex set containing RðGÞÞ:We will show below that these four systems can indeed be defined,in other words,Co RðGÞ;Z.-B.Xu et al./Neural Networks17(2004)73–85 74RðW qÞand RðWGÞþq do form invariant sets or manifolds of Eqs.(3)and(4),respectively.We will assume throughout this paper the existence of aunique solution xðt;x0Þto Eq.(3)and a unique solutionyðt;y0Þto Eq.(4)for any given initial states x0;y0in R N (which is the case, e.g.when G is locally Lipschitzcontinuous).As usual,the solution xðt;x0Þis also called atrajectory of Eq.(3)through x0denoted henceforth by G1ðx0Þ(correspondingly,the trajectory yðt;y0Þof Eq.(4) through y0is denoted by G2ðy0Þ).A subset D,R N is said to be an invariant set of the system(3)if x0[D implies G1ðx0Þ#D:Wefirst clarify the trajectory transformation relationshipbetween systems(3)and(4.1).Theorem 1.Trajectory transformation relationship.IfG1ðx0Þis the trajectory of Eq.(3)through x0[R N;then W G1ðx0Þþq is the trajectory of Eq.(4.1)through y0¼Wx0þq;that is,G2ðWx0þqÞ¼W G1ðx0Þþq:Conversely,if yðtÞ¼G2ðWx0þqÞis the trajectory of Eq.(4.1)through y0¼Wx0þq for some x0[R N;then xðt;x0Þ; defined byxðt;x0Þ¼e2t=t x0þ1te2t=tðte s=t GðyðsÞÞd s;ð5Þis the trajectory of Eq.(3)through x0;that is,G1ðx0Þ¼xðt;x0Þ:Proof.First,let G1ðx0Þ¼xðt;x0Þ:Then xðt;x0Þsolves the Eq.(3),that is,t d xðt;x0Þd t¼2xðt;x0ÞþGðWxðt;x0ÞþqÞ:Multiplying both sides of this equation with W gives the resultt d½Wxðt;x0Þþqd t¼2Wxðt;x0Þ2qþWGðWxðt;x0ÞþqÞþq;which means thatt d yðt;y0Þd t¼2yðt;y0ÞþWGðyðt;y0ÞÞþq;where yðt;y0Þ¼Wxðt;x0Þþq and y0¼Wx0þq:Thus G2ðWx0þqÞ¼W G1ðx0Þþq is the trajectory of Eq.(4.1) through Wx0þq:Now if yðtÞ¼G2ðWx0þqÞis the trajectory of Eq.(4.1) through y0¼Wx0þq;then we havet d yðtÞd t¼2yðtÞþWGðyðtÞÞþq:This leads toyðtÞ¼e2t=t yð0Þþ1te2t=tðte s=t WGðyðsÞÞd sþð12e2t=tÞq¼e2t=t½yð0Þ2q þ1e2t=t Wðte s=t GðyðsÞÞd sþq¼W e2t=t x0þ1te2t=tðte s=t GðyðsÞÞd sþq¼Wxðt;x0Þþq;ð6Þwhere xðt;x0Þis defined as in Eq.(5).On the other hand,adirect calculation using Eq.(5)shows thattd xðt;x0Þd t¼2e2t=t x0þGðyðtÞÞ21te2t=tðte s=t GðyðsÞÞd s¼2xðt;x0ÞþGðyðtÞÞ:From Eq.(6)it then follows thattd xðt;x0Þd t¼2xðt;x0ÞþGðWxðt;x0ÞþqÞ:That is,xðt;x0Þis the trajectory of Eq.(3)through x0:TheProof of Theorem1is thus complete.ATheorem1implies that the trajectories of systems(3)and(4.1)can be transferred from one to the other.However,itdoes not mean that the trajectories of Eq.(3)and(4)can betransferable unless W is invertible.This is because,in thecase when W is not invertible,RðW qÞis a strict subset of R Nso any trajectory G2ðy0Þof Eq.(4)may not be expressible interms of the solution of Eq.(3).This of course causesdifficulty in our subsequent discussion.However,it isadequate for us to apply Theorem1to establish thefollowing result which implies that systems(3.1),(3.2),(4.1)and(4.2)can indeed be well-defined.Theorem2.Invariance-manifold property(i)Any closed convex manifold S;containing RðGW qÞ(orRðGÞ),is an invariant manifold for system(3).(ii)RðW qÞis an invariant manifold for Eq.(4).(iii)RðWGÞþq is an invariant manifold for Eq.(4)provideRðGÞis bounded and convex.Proof.(i)Let x¼xðt;x0Þbe a solution of Eq.(3)startingfrom x0[S and let LðxÞ¼Wxþq:Then,by the theory ofdifferential equations(see,e.g.Verhulst,1990),we havexðt;x0Þ¼e2t=t xð0Þþ1te2t=tðte s=t GðLðxðsÞÞÞd s¼e2t=t x0þð12e2t=tÞe2t=tðte s=t GðLðxðsÞÞÞd sð12e2t=tÞtð7ÞZ.-B.Xu et al./Neural Networks17(2004)73–8575WriteIðtÞ¼e2t=tðte s=t GðLðxðsÞÞÞd sð12e2t=tÞtThen Eq.(7)implies that,since e2t=t[ð0;1Þand S is a convex set,xðt;x0Þ[S if IðtÞ[S:Thus in order to prove(i) it is enough to prove that IðtÞ[S:Note thatð12e2t=tÞt¼ðte2s=t d s¼lim1!1X nk¼1e2j k=t D t kande2t=t ðte s=t GðLðxðsÞÞÞd s¼ðte2ðt2sÞ=t GðLðxðsÞÞÞd s¼ðte2r=t GðLðxðt2rÞÞÞd r¼limn!1X nk¼1e2j k=t GðLðxðt2j kÞÞÞD t k:So IðtÞ¼lim n!1I nðtÞwithI nðtÞ¼X nk¼1ðe2j k=t D t kÞGðLðxðt2j kÞÞÞX nk¼1e2j k=t D t k:Now letl k¼e2j k=t D t kX nk¼1e2j k=t D t k:Then l k[½0;1 and P nk¼1l k¼1so,I nðtÞ[S for any nsince GðLðxðt2j kÞÞÞ[RðGÞ#S by the assumption and S is convex.Thus IðtÞ[S since S is closed,which completes the proof of(i).(ii)Let P¼RðWÞþq:Then for y0[P there is an x0[ R N such that y0¼Wx0þq:Let xðt;x0Þbe the solution of Eq.(3)through x0:Then it is easy to verify that yðt;y0Þ:¼Wxðt;x0Þþq is a solution of Eq.(4)through y0:Clearly, yðt;y0Þ[RðWÞþq:This proves(ii).(iii)If RðGÞis a bounded convex set,then Co RðGÞ¼RðGÞ: Taking S¼RðGÞand arguing as in the proof of(i)we obtain that G2ðy0Þ[WðRðGÞÞþq whenever y0[ WðRðGÞÞþq:To complete the proof we need to show that WðRðGÞÞ¼RðWGÞby using the fact that W is linear and RðGÞis bounded.First,it is clear that WðRðGÞÞ# RðWGÞ:Now if x[RðWGÞ;then there is a sequence {WGðx jÞ}such that lim j!1WGðx jÞ¼x:Since{Gðx jÞ}is bounded,we may assume without loss of generality that lim j!1Gðx jÞ¼x p for some x p[From the linearity of W it follows that x¼lim j!1WGðx jÞ¼Wx p;which implies that x[WðRðGÞÞ:So RðWGÞ#WðRðGÞÞ:Conse-quently,WðRðGÞÞ¼RðWGÞ;which completes the proof of Theorem2.A Remark1.(i)From the Proof of Theorem2it is easily seen that Theorem2(i)is actually true for any generic nonlinear operator L;not necessarily restricted to the affine transform-ation case as defined in Eqs.(3)and(4).This may manifest an exclusive feature of the static neural network model(3). (ii)Theorem2says that any solutions xðt;x0Þand yðt;y0Þwill remain in the manifolds Co RðGÞ;Co RðGW qÞ;RðW qÞand WðRðGÞÞþq if they are initialized from them.It says, however,nothing about what would happen if the solutions are not starting from the manifolds.Theorem3provides us with some information in this case.Theorem3.Attractive manifold property(i)For any x0[R N the trajectory G1ðx0Þexponentially approaches to Co RðGÞ(or Co RðGW qÞÞin the sense that dðxðt;x0Þ;Co RðGÞÞ#e2t=t dðx0;Co RðGÞÞ;ð8Þwhere dðx;SÞ¼inf y[S k x2y k denotes the distance of x to the set S:(ii)For any y0[R N the trajectory G2ðy0Þexponentially approaches to RðW qÞ(or RðWGÞþq in the case when RðGÞis bounded and convex)in the sense thatdðyðt;y0Þ;RðWÞþqÞ#e2t=t dðy0;RðWÞþqÞ:ð9ÞProof.(i)It follows from Eq.(7)that for any x1[Co RðGÞ(or Co RðGW qÞ),xðt;x0Þ¼e2t=tðx02x1Þþe2t=t x1þ1e2t=tðte s=t GðLðxðsÞÞÞd s: LetzðtÞ¼e2t=t x1þ1te2t=tðte s=t GðLðxðsÞÞÞd s:Then argue as in the Proof of Theorem2to obtain that zðtÞ[Co RðGÞ(or Co RðGW qÞ)for all t$0:Thus,dðxðt;x0Þ;Co RðGÞÞ#k xðt;x0Þ2zðtÞk#e2t=t k x12x0kfor any x1[Co RðGÞ;which implies Eq.(8).(ii)For any y1[RðWÞþq there is an x1[R N such that y1¼Wx1þq:Thus we haveyðt;x0Þ¼e2t=tðy02y1Þþe2t=t y1þ1te2t=tðte s=tðWGðyðsÞÞþqÞd s¼e2t=tðy02Wx12qÞþe2t=tðWx1þqÞþ1e2t=tðte s=tðWGðyðsÞÞþqÞd sV e2t=tðy02Wx12qÞþzðtÞ:Similar argument as in the Proof of Theorem2together with the convexity and closedness of RðW qÞ(or RðWGÞþq) gives that zðtÞ[RðW qÞ(or RðWGÞþq).Thus it followsZ.-B.Xu et al./Neural Networks17(2004)73–85 76thatdðyðt;y0Þ;RðW qÞÞ#k yðt;x0Þ2zðtÞk#e2t=t k y02y1k:This implies Eq.(9)since y1[RðW qÞis arbitrary.The Proof of Theorem3is thus complete.ARemark2.Theorem3implies that Co RðGÞ;Co RðGW qÞand RðW qÞ(or RðWGÞþq in the case when RðGÞis bounded and convex)are,respectively,the superior nontrivial attractive manifold of systems(3)and(4).2.2.Equilibrium relationshipsIn this section we will study the relationship among the equilibria of systems(3),(4),(3.1),(3.2),(4.1)and(4.2).Notefirst that x p is an equilibrium state of system(3)if and only if it is an equilibrium state of Eqs.(3.1)and(3.2) and that x p solves the equation F1ðxÞ¼0;that is,x p is a zero of F1;where F1is defined byF1ðxÞ¼2xþGðWxþqÞ;;x[R N:ð10ÞSimilarly,y p is an equilibrium state of Eq.(4)if and only if it is an equilibrium state of Eqs.(4.1)and(4.2)and further,y p is a zero of F2with F2being defined byF2ðyÞ¼2yþWGðyÞþq;;y[R N:ð11ÞDenote by F211ð0Þthe common equilibrium state set of Eqs.(3),(3.1)and(3.2)and by F212ð0Þthe equilibrium state set of Eqs.(4),(4.1)and(4.2).Theorem4.Equilibrium relationshipx p[F211ð0Þif and only if y p¼Wx pþq[F212ð0Þ:Con-versely,y p[F212ð0Þif and only if x p¼Gðy pÞ[F211ð0Þ: Proof.If x p[F211ð0Þis an equilibrium state of Eq.(3),then x p¼GðWx pþqÞ¼Gðy pÞ:Multiplying both sides with W then gives Wx p¼WGðy pÞ:So y p¼WGðy pÞþq;which means that y p[F212ð0Þand is an equilibrium state of Eq.(4).On the other hand,if y p[F212ð0Þis an equilibrium state of Eq.(4),then y p¼WGðy pÞþq:Let x p¼Gðy pÞ:Then x p¼Gðy pÞ¼GðWGðy pÞþqÞ¼GðWx pþqÞ:So x p¼Gðy pÞ[F211ð0Þand is an equilibrium state of Eq.(3).The proof is thus complete.AFrom Theorem4it seems that systems(3)and(4)should have the same number of equilibria.This is,however,by no means evident if W is singular and G is not invertible. Theorem5ensures that such a conclusion is actually true being independent of the nonsingularity of W and the invertibility of G:Theorem5.Identical cardinal number of equilibria.The numbers of equilibrium states of systems(3)and(4)are identical.Proof.Let CardðF211ð0ÞÞand CardðF212ð0ÞÞdenote the cardinal numbers of the equilibrium state sets F211ð0Þand F212ð0Þ;respectively.We will show that CardðF211ð0ÞÞ¼CardðF212ð0ÞÞby proving that there exists a bijective mapping between F211ð0Þand F212ð0Þ:In fact,consider the mapping P:F212ð0Þ!F211ð0Þdefined byF212ð0Þ]y p!Gðy pÞ¼x p[F211ð0Þ:By Theorem4,G is well-defined.First,G is injective since, if Pðy p1Þ¼Pðy p2Þfor some y p1;y p2[F212ð0Þthen,by the definition of equilibrium state of Eq.(4),y p i¼WGðy p iÞþq (i¼1;2)so that y p1¼WGðy p1Þþq¼WGðy p2Þþq¼y p2: Finally,P is surjective since,for any x p[F211ð0Þ;we have x p¼GðWx pþqÞso,if y p¼Wx pþq then,by Theorem4,y p[F212ð0Þ;that is,x p¼Gðy pÞ:This proves the theorem.AFrom the Proof of Theorem5it follows that the nonlinear activation mapping G;when restricted to the equilibrium state set F211ð0Þ;is invertible no matter whether or not the activation mapping G itself is invertible.In this case,the inverse G21can be given byG21ðx pÞ¼Wx pþq;;x p[F211ð0Þ:ð12ÞSimilarly,we can conclude that W;when restricted to F212ð0Þ2q;is regular even though W itself may not be regular.Thus,x p[F211ð0Þif and only if y p¼G21ðx pÞ[ F212ð0Þ:Conversely,y p[F212ð0Þif and only if x p¼Gðy pÞ[F211ð0Þ:A pair of equilibria x p and y p possessing such a property will be called as a pair of mutually mapped equilibria of system(3)and(4),denoted henceforth by ðx p;y pÞ:With such understanding,we state and prove a series of stability–invariance properties of systems(3)and (4.1)in Section2.3.2.3.Stability–invariance propertiesWefirst recall some notion and notations from the theory of dynamical systems(taking Eq.(3)as an example).A point x p is said to be a v-limit point of G1ðx0Þif there is a subsequence{t i}such that x p¼lim i!1xðt i;x0Þ:All the v-limit points of G1ðx0Þconstitute the v-limit set of G1ðx0Þ; denoted henceforth by vðG1ðx0ÞÞ:The v-limit set is invariant under the dynamics.The equilibrium point x p is said to be stable if any trajectory of Eq.(3)can stay within a small neighborhood of x p whenever the initial state x0is close to x p:The equilibrium point x p is said to be attractive if there is a neighborhood Jðx pÞ;called the attraction basin of x p;such that any trajectory of Eq.(3)initialized from a state in Jðx pÞwill approach to x p as time goes to infinity.An equilibrium state is said to be asymptotically stable if it is both stable and attractive.Further,the equilibrium state x p is said to be exponentially stable if there exist a constant a. 0and a strictly increasing function M such thatk xðt;xÞ2x p k#Mðk x02x p kÞe2a t:ð13ÞZ.-B.Xu et al./Neural Networks17(2004)73–8577If,in addition,MðsÞ¼Ks(i.e.M is linear)for an absolute constant K in Eq.(13)then x p is said to be exponentially stable in the sense of Liapunov.The equilibrium state x p is globally asymptotically stable(globally exponentially stable)if it is asymptotically stable(exponentially stable) and Jðx pÞ¼R N:Correspondingly to the global stability notion,we say that a system(say,Eq.(3))is globally convergent if xðt;x0Þconverges to an equilibrium state of Eq.(3)for every initial point x0[R N:Note that the limit of xðt;x0Þmay be different for different x0:The system(3)is said to be exponentially convergent if it is globally convergent with the limiting state x p satisfying Eq.(13).We need the following two simple lemmas.Lemma1.Let W be an N£N matrix and let B be the unit open ball of R N centered at zero and with radius1,i.e. B¼{x lk x k,1}:Set WðBÞ¼{Wx l x[B}:Then there is a constant d.0such thatd1B>RðWÞ#WðBÞ;whenever d1#d:Proof.It is clear that WðBÞis a neighborhood of0in the subspace RðWÞ,R N:Thus there is an open set U in the space RðWÞsuch that0[U#WðBÞ:Since U is a relatively open set of R N;then there must be an open set V in R N such that U¼V>RðWÞ:Clearly,V is a neighborhood of0in R N:Thus there is a constant d.0 such that,whenever d1#d;we have d1B>RðWÞ# V>RðWÞ#WðBÞ:This proves the lemma.ALemma2.If RðGÞis bounded and convex,then RðWGÞþq¼WðRðGÞÞþq:Proof.The inclusion WðRðGÞÞþq#RðWGÞþq is trivial. To prove the inverse inclusion,take anyfixed y[ RðWGÞþq and assume that{y i}#R N is a sequence such that y¼lim i!1WGðy iÞþq:Let x i¼Gðy iÞ:Then x i[ RðGÞ:Since RðGÞis bounded,we may assume without loss of generality that lim i!1x i¼lim i!1Gðy iÞ¼x for some x[RðGÞ:It follows that y¼lim i!1WGðy iÞþq¼lim i!1Wx iþq¼Wxþq:This means that y[WðRðGÞÞso RðWGÞ#WðRðGÞÞ:Consequently,WðRðGÞÞþq¼RðWGÞþq:The lemma is thus proved.AWith the above lemmas,we are now in a position to examine the relationship among the stability of models(3), (4),(3.1),(3.2),(4.1)and(4.2).Theorem6.Stability.Letðx p;y pÞ[F211ð0Þ£F212ð0Þbe any pair of mutually mapped equilibria of systems(3)and (4).Assume that G is a Lipschitzian,that is,there is a positive constant LðGÞsuch thatk Gðy1Þ2Gðy2Þk#LðGÞk y12y2k;;y1;y2[R N:Then x p;as an equilibrium state of Eq.(3),is stable if andonly if y p;as an equilibrium state of Eq.(4.1),is stable.Proof.‘):’If x p[F211ð0Þis stable,then for any1.0there is d1ð1Þ.0such that,whenever k x02x p k,d1ðeÞ,1;we havek xðt;xÞ2x p k,k W k211;t$0:ð14ÞLet d,d1ð1Þbe so small that d B>RðWÞ#d1ðeÞWðBÞ:(By Lemma1this is possible.)We now prove that for y0[RðWÞþq;k yðt;yÞ2y p k,1whenever k y02y p k,dð1Þ;which together with Lemma2implies that y p;as an equilibrium state of system(4.1),isstable.For any y0[RðWÞþq and y p[F212ð0Þwe have y02y p[RðWÞ;so y02y p[d1ð1ÞWðBÞwhenever k y02y p k,dð1Þ:Thus there is a z0[d1ð1ÞWðBÞso that we can writey02y p¼Wz0or equivalently,y0¼Wðz0þx pÞþq:FromTheorem1it follows that if xðtÞis the solution of system(3)through x0¼z0þx p then yðt;y0Þ¼WxðtÞþq is the uniquesolution of Eq.(4.1)passing through y0:So,for any t.0;we havek yðt;yÞ2y p k¼k WxðtÞ2Wx p k#k W kk xðtÞ2x p k:This,together with Eq.(14),and on noting that k x02x p k¼k zk,d1ð1Þ;implies that k yðt;y0Þ2y p k,1;as expected.‘(:’Given any1.0;by assumption there is a d1ð1Þ.0such that for y0[RðWÞþq;we have k yðt;y0Þ2y p k,1=½2LðGÞ for t$0whenever k y02y p k,d1ð1Þ:Lettingdð1Þ¼mind1ð1Þk W k;12;we can prove that for x0[R N;k xðt;x0Þ2x p k,e providedk x2x p k,dð1Þ:In fact,let y0¼Wx0þq:Then y0[RðWÞþq and satisfies that k y02y p k,d1ð1Þwheneverk x2x p k,dð1Þ:So,by Theorem1,we can write yðt;y0Þ¼Wxðt;x0Þþq:Further,we havek yðt;yÞ2y p k,1whenever k x02x p k,dð1Þ:From Eq.(5)it thus follows thatk xðt;xÞ2x p k#e2t=t k x02x p kþ1te2t=tðte s=t GðyðsÞÞd s2ð12e2t=tÞx p¼e2t=t k x02x p kþ1te2t=tðte s=t½GðyðsÞÞ2Gðy pÞ d s,e2þe21te2t=tðte s=t d s,1:This implies the stability of x p since1is arbitrary.Thetheorem is thus proved.AZ.-B.Xu et al./Neural Networks17(2004)73–8578。