A trace formula for the forcing relation of braids

合集下载

TRANSLATOR

TRANSLATOR

DESIGN AND IMPLEMENTATION OF SIND,A DYNAMIC BINARYTRANSLATORbyTREK PALMERputer Science,University of New Mexico,2001M.S.,Computer Science,University of New Mexico,2003AbstractRecent work with dynamic optimization in platform independent,virtual machine based languages such as Java has sparked interest in the possibility of applying similar techniques to arbitrary compiled binary programs.Systems such as Dynamo,DAISY, and FX32exploit dynamic optimization techniques to improve performance of native or foreign architecture binaries.However,research in this area is complicated by the lack of openly licensed,freely available,and platform-independent experimental frameworks. SIND aims tofill this void by providing an easily-extensible andflexible framework for research and development of applications and techniques of binary translation.Current research focuses are dynamic optimization of running binaries and dynamic security aug-mentation and integrity assurance.ContentsList of Figures v List of Tables vi 1Introduction11.1Dynamic Binary Translation (2)1.2Why Another One? (3)1.3SIND (3)1.4Overview of the Thesis (4)2Previous Efforts52.1Dynamo (5)2.2DynamoRIO (7)2.3FX32 (8)2.4DAISY (8)2.5Crusoe,JVMs,and Others (9)3SIND Design103.1Interpreter (12)3.1.1Registers (13)3.1.2Instructions (14)3.1.3Exceptional Conditions (15)3.1.4Signals and Asynchronous I/O (15)3.2Memory Manager (16)3.3Syscall Manager (17)3.4Trace Gathering (18)3.5Transformers (19)3.6Fragment Cache (19)3.7Bootstrapper and Dispatcher (20)3.7.1Bootstrapper (20)3.7.2Dispatcher (21)4SIND Implementation224.1Overview (22)4.2Interpreter (23)4.3Bootstrapper (24)4.4Fragment Cache (25)5Evaluation of SIND265.1Performance of SIND (26)5.1.1Speed of Interpretation (26)5.1.2Speed of Cached Fragments (28)5.2Memory Footprint of SIND (28)5.3The Agility of SIND (28)6SIND:A History306.1A Brief History (30)6.2Experiences from the Design and Implementation of SIND (32)7Using SIND367.1Invoking SIND (36)7.2Extending SIND (37)7.2.1System-Dependent Code (37)7.2.2System-Independent Code (38)A Technical Details39A.1Source Layout and Directory Organization (39)A.2Where Functionality Resides (40)A.3Compilation and Architecture Support (41)A.3.1Makefiles (41)A.3.2Compilation Flags (41)A.3.3Supported Architectures (42)References43List of Figures3.1SIND modules (11)4.1CPU inheritance tree (23)List of Tables5.1SIND interpreter slowdown (27)A.1Includefiles to modules (40)A.2modules’sourcefiles (41)Chapter1IntroductionProgram transformation and optimization are not new ideas;however the notion of per-forming them at runtime is a recent invention.The attraction of dynamic transformation and analysis of programs derives,in part,from the fact that the amount of static informa-tion available to a compiler is shrinking.Object-oriented languages that support dynamic loading and unloading of code impose a serious restriction on the ability of static analysis to effectively guide optimization.Even in a fairly static O-O language such as C++,there are considerably more challenges for a static compiler to overcome than in C.Static com-pilation techniques fail primarily because they have insufficient information to work with. When statically compiling a program with loadable modules,for instance,the compiler cannot make many assumptions about the internal structure of those modules,and conse-quently standard optimization techniques(such as inlining)become impossible.Also,for static analysis to be fully effective,it is often necessary to have access to the source code. For instance,some of the most effective security analysis programs(such as StackGaurd [4])need full access to the source code to correctly protect vulnerable code segments.In reality,however,source code is often impossible to obtain.1.1Dynamic Binary TranslationDynamic binary translation is a method to overcome these deficiencies in static compi-lation techniques.The basic idea is to dynamically monitor a running program and use the gathered profile(which contains a great deal of information)to guide further program transformations.The challenge is to do this efficiently and transparently.Both efficiency and transparency are difficult problems.To provide transparency,the binary translator must emulate all the idiosyncrasies of the underlying system,and do so in such a way that control never leaves the translator.This is the problem that most debuggers have to solve. However,whereas a debugger developer can choose to sacrifice speed for convenience and reliability,such trade-offs cannot be made with a dynamic binary translator.The efficiency constraint means that the system as a whole must be able to simulate the underlying archi-tecture without significantly slowing down the running program.In practice,this means about a10%slowdown is acceptable.This leads to the use of many arcane tricks and techniques to achieve the seemingly impossible goal of dynamic profiling with almost no slowdown.These are what makes designing and implementing a dynamic binary translator such an engineering challenge.Most binary translators have the same basic components:something to profile the running code and gather traces,something to transform the traces into fragments,and something to link the fragments up with the program’s address space and run the frag-ments directly on the processor.It is the last step that makes up for the cost of profiling and transforming running code.Because most of a program’s execution is confined to a small portion of the code,if that portion is optimized and run directly on the proces-sor it would speed up execution significantly.Identifying those hot sections of the code, however,is effectively impossible to do statically.Fortunately,the runtime characteristics of many programs are often much simpler than the static whole-program characteristics. This simplicity,along with the increased information available at runtime,makes it pos-sible for dynamic translators to identify important segments of code that static analysiswould not be able to catch.This ability can be used to guide instance-tailored optimiza-tions to improve program performance,or to guide runtime security transformations to improve program stability and security.1.2Why Another One?While SIND is not thefirst attempt at a dynamic binary translator,it is thefirst open one. Many previous systems were created by companies as either an internal research tool[1]or as a commercial product[3],and consequently were never released.As the discussion in Chapter2reveals,many of these systems also have deficiencies or peculiar requirements. Some systems are so tied to the target platform that porting them was infeasible[1].Other systems have specific(and unobtainable)hardware requirements[5].SIND was designed with all this in mind.The SIND framework was intentionally constructed to be portable and the current implementation for the UltraSPARC can run on commodity hardware with no custom components.These advantages alone warrant the development of SIND.In addition,SIND is an open-source system and as such may become an important research tool with a large developer base.It was,in fact,the very lack of such an open tool that motivated the development of SIND.1.3SINDSIND is my effort at designing a platform-independent dynamic binary translation frame-work,and implementing that framework for the UltraSPARC architecture(running So-laris/SPARC).This effort stems from the fact that there are no real open platforms for doing dynamic translation.This hinders research,especially in afield where open re-search tools are the norm.SIND’s design was abstracted from several published dynamic binary translators and is aimed at providing a general framework for building a binarytranslator for any given platform.The whole system is organized in an O-O fashion,with each major component as a separate module.This modularity is intended to aid initial and subsequent development.In the future,it should be possible to extend a module without having to modify any other code.The current SIND system implements user-level integer instructions.No supervisor or floating point code is currently supported.The lack of supervisor code is not a problem because in a modern UNIX-like system(such as Solaris),all the supervisor code lives in kernel space.Requests to this code are made through syscalls,which are proxied by SIND. Floating-point code is less common than integer code,but for SIND to be truly useful it must handlefloating point operations.These instructions were intentionally passed over due to the potential complexity of implementing them correctly(and the resultant debug-ging nightmare).SIND is also unaware of the Solaris threading infrastructure and may therefore be insufficiently thread-safe.1.4Overview of the ThesisThe remainder of this document is organized as follows:Chapter2gives a summary of the previous efforts at dynamic binary translation;Chapter3describes,in detail,the design of the SIND system;Chapter4is a discussion of the current implementation of SIND for the UltraSPARC architecture;Chapter5is a discussion and evaluation of the performance andflexibility of the current experimental system;Chapter6is an overview of the history of the project as well as a discussion of large scale technical issues encountered while implementing SIND;Chapter7describes how to use the developmental SIND tool;and lastly the document is concluded with an appendix describing the code itself and details such as directory structure.Chapter2Previous EffortsThere have been several notable efforts at dynamic binary translation.The most well-known is the Dynamo project from HPLabs.This was a dynamic optimization research project for HPPA systems running HP-UX.Another interesting project was the FX32 project from DEC(now part of HP/Compaq).This system was a dynamic binary translator that ran IA32binaries on an Alpha(running Windows NT).FX32had several notable fea-tures:not only did it efficiently transform foreign binary instructions,it persistently stored the fragments on disk and optimized them in an offline batch-processing phase.Other interesting systems include the DAISY project from IBM,the Hotspot and Jalape˜n o/Jikes JVMs,and Transmeta’s Crusoe‘code-morphing’technology.2.1DynamoThe Dynamo system[1]is,in many ways,the seminal effort in thisfield.It is the most popularized effort that actually achieves noticeable improvements in running time.The Dynamo system is geared toward dynamic runtime optimization of HPPA binaries run-ning under a custom version of HP-UX.The system is bootstrapped by a hacked version ofcrt.o and begins running the binary immediately.The instructions are fully interpreted by a software interpreter,whose primary task is to identify and capture hot code traces from the running program.The profiling method used is one of the Dynamo team’s fun-damental contributions.They experimented with several profiling metrics and found that a simple statistical approach yielded the best combination of accuracy and speed.First, the profiler only focuses on code traces that started with trace heads,namely backwards-taken branches.These branches are indicative of loops within the program,and Dynamo assumes that this is where most of a program’s work gets done.Secondly,the profiler assumes that,on average,the branches being taken when it examines the code would be the ones the program would normally take.Therefore when a trace head became hot(was visited enough times),only a single code trace would be gathered.This code trace is then run through several simple compiler passes to yield an opti-mized fragment.Because overhead had to be small the compiler only performs simple, linear pass optimizations.The fragments are then loaded into the fragment cache.Dy-namo’s cache holds the other fundamental contribution.Rather than just linking the frag-ment so that it correctly accessed program data,the fragment is also potentially linked to other fragments already in the cache.This obviates the need to leave the fragment cache from one fragment merely to have to re-enter the cache to execute another fragment.This single improvement led to impressive performance increases.Despite its many successes Dynamo has many disadvantages.First,the system is not open.This seriously hampers research,as the tool cannot be extended when necessary. Second,Dynamo is specifically tailored to the HPPA architecture and HP-UX operating system,to which I do not have access.Thirdly,Dynamo would be difficult to port,even if the source code was publicly available,the system wasn’t engineered to be particularly extensible.The HP engineers who wrote Dynamo admitted that the whole system may have to be rewritten to be useful on another platform.2.2DynamoRIODynamoRIO[2]is the successor to Dynamo.It is also a closed,proprietary system,but it is designed for the Intel IA32(x86)architecture and has versions that run under Win-dows and Linux.In addition to the standard problems of building a dynamic optimization system,DynamoRIO had to overcome the enormous cost of interpreting the dense and complex x86instruction set.After several false starts,this was eventually achieved with the use of a so-called basic-block cache.This is a form of‘cut&paste’interpretation in which the interpreter/decoder fetches basic blocks from memory,rewrites branch/jump targets and executes the modified code directly on the processor.This alleviated the dif-ficulty of actually interpreting x86instructions,but made profiling more complex.The initial decision to use a basic-block cache also tied the rest of the system to the x86archi-tecture1.In the cause of efficiency,each component of DynamoRIO was written to be as specific to the x86architecture as possible.As a consequence,the entire system is highly non-portable and would have to be completely rewritten to handle a new instruction set [Personal Discussions with Derek Bruening,the DynamoRIO Maintainer].Despite the tight coupling between DynamoRIO and the x86architecture,the system is more open andflexible than the original Dynamo.Even with the closed nature of the underlying source code,there is a useful API that allows outside developers to add to the system.However,such outside additions are restricted by the API and are slowed by the need to pass data through an additional interface not used by DynamoRIO internals. Despite the improvements over the original Dynamo,DynamoRIO failed to fully solve the portability and extensibility problems.2.3FX32FX32[3]is a dynamic translation program from DEC.It is designed to translate IA32 binaries to Alpha code at runtime.The whole system runs on Windows NT for the Alpha, and existed because many NT developers were either unable or unwilling to write Alpha-friendly code.FX32has a number of notable features.It does a great job of translating foreign binaries,facilitated primarily by the fact that the NT API is standard across both the Alpha and IA32platforms.This allows rapid translation of system and library calls in a1-to-1fashion.FX32also optimizes the translated traces,but in a novel way.Rather than doing optimizations at runtime the FX32system simply translates the trace and then saved the translated version to ter on,a batch job examines the saved traces and optimizes them using potentially long-running algorithms.In practice,this means that each time a user ran an IA32application it would be somewhat faster than the time before.FX32is an interesting piece of software,but it too suffers from serious drawbacks. Primarily,it suffers from the fact that its a closed-source system.DEC(and later Compaq) sold FX32along with NT for Alpha,and considered releasing the code to be economically impossible.Also,FX32is closely tied to the NT platform,which can be difficult to develop for.2.4DAISYDAISY[5]is a binary translation project from IBM that performs dynamic compilation of Power4binaries.It is similar to Dynamo in principle,but it employs more sophisticated translation and profiling schemes.This allows DAISY to do a more sophisticated analysis than Dynamo.For instance,a limited form of controlflow analysis across branches and calls is performed(to eliminate as much indirection as possible).However,this added power comes at the cost of larger runtime overhead.The DAISY project obviated thiscost by creating a custom daughter-board that would house an auxiliary processor to run the DAISY system.This secondary processor only has to run at a fraction of the speed of the main processor,and the daughter board has several megabytes of isolated memory available only to the auxiliary processor.Because of this,DAISY has automatic memory protection at no runtime cost,the additional hardware also removes the distinction between the operating system and user applications.This means that DAISY can optimize both OS code and application code(and even optimize call sequences from one through the other).Unfortunately,the DAISY project never produced a commercially-available version of the DAISY processor in hardware.All the published results came from detailed soft-ware simulation of the proposed hardware.Even if the hardware were eventually mass-produced,it was intended for use in high-end servers,and so would probably have been very expensive.2.5Crusoe,JVMs,and OthersOther dynamic translation projects include the Code-morphing technology used in Trans-meta’s Crusoe processor[7],the HotSpot and Jalape˜n o/Jikes optimizing JIT JVMs,and other virtual machines that employ dynamic(otherwise known as Just In Time)compila-tion techniques.The main disadvantage with code-morphing is that in addition to being proprietary,it is specifically tied to the Crusoe VLIW architecture.The JVM and other language virtual machine projects,although useful from a design perspective,did not con-tribute much to the actual construction of SIND.This is due to the virtual machines being tailored to the needs of a specific language.This means that most language virtual ma-chines,although closer to hardware than the uncompiled program,have features useful to the source language that are difficult to map directly to hardware.Chapter3SIND DesignThe SIND system design is not particularly revolutionary.It is a synthesis and extension of many dynamic translator designs.Because most dynamic binary translators have to solve similar problems,many have similar designs.This similarity is encouraging,because it means that if this structure can be expressed in code,the construction of new binary translators would be reduced to extending the base modules,rather than designing the whole system from scratch.Figure3.1shows the major components of SIND.The interpreter is the module that handles the dynamic execution and profiling of the running binary.The transformers trans-late gathered traces into fragments.The fragment cache handles fragment linking and runs the fragment code on the processor.The memory and syscall-manager handle the sys-tem specific aspects of memory protection and operating system interaction,respectively. They are separated from the other modules to ensure as much platform independence as stly,the bootstrapper and dispatcher initialize the other modules and handle inter-module communication.This framework is generic enough to encompass all source and target architecture con-figurations,and separates the components so that they may act like‘plug-in’modules.Figure3.1:SIND modulesFor instance,because the interpreter accesses memory through the Memory Manager and accesses OS functionality through the Syscall Manager,the interpreter only has to emu-late the source architecture and has no dependence on the operating system.This means, ideally,that an UltraSPARC interpreter would be able to run(without modifying the in-terpreter source)on both an UltraSPARC and Power4and would trust the Memory and Syscall Managers to take care of OS specifics.The intention is to isolate the interpreter from all but the most large-scale details of the target architecture.Basically,only the endi-aness and bit-width of the underlying system need to be taken into account(and this,only because C++specifies no standards for the size and endianess of data).The basic operation of the SIND framework is also platform-independent.The pro-gram to be run under dynamic translation is started by the bootstrapper.The bootstrap-per assures that all dependencies(libraries and other shared objects)are loaded and halts execution just before the program starts.Then,control passes to the dispatcher,which initializes all the remaining SIND modules and starts up the interpreter.The interpreterdynamically executes the program and gathers profiling information.According to some internal metric,the interpreter eventually decides that it has encountered an interesting code segment and gathers the relevant instructions and processor context into a trace.This trace is then handed(through dispatch)to the transformers.The transformers transform the trace into a functionally equivalent fragment.The nature of the transformations could be varied.Traces could be rewritten to be more efficient,but they could also be rewritten to be more secure,or to generate morefine-grained profiles.When thefinal transformer hasfinished its transformations,the fragment is handed to the fragment cache.The cache’s primary responsibilities are to guarantee that the running fragment will have transparent access to all program data,and to simultaneously guarantee that the running fragment will not modify SIND or break out of SIND.The cache can protect SIND data by selectively write-protecting the regions of memory that SIND inhabits when the cache is entered and un-write-protecting the regions when the cache is exited.The cache also needs to check for system calls that might un-protect the SIND memory regions.It can do this by placing itself between the executing fragment and the eventual system call,and checking on the parameters passed in by the fragment.The cache can guarantee that an executing fragment will be able to access all program data by performing afinal rewriting of the fragment in a process analogous to dynamic linking and loading.From then on,when program control reaches the address of a fragment,control is passed to the cache,and the fragment then runs directly on the processor.When control leaves the fragments in the cache,the SIND interpreter starts up again and continues dynamic program execution.3.1InterpreterThe interpreter’s main function is to gather profiling information and code execution traces.These are passed to transformers,which use the profiling information to guide specialized transformations of the code traces.Because one of the goals of the SIND sys-tem is to do runtime binary optimization,it is vital that the interpreter should introduce as little overhead as possible.As a consequence,the interpreter must be very efficient and every reasonable effort must be made to improve its speed.Thefirst interpreter to be fully designed and implemented in SIND emulates the64-bit SPARC v9architecture.The design was motivated by several factors:first,because SIND runs in non-privileged mode,the interpreter is primarily a non-privileged instruc-tion interpreter;second,the interpreter only needs to be functionally correct,therefore no complicated hardware structure needs to be emulated in order to produce accurate simu-lation.The interpreter’s job is then to replicate a user’s view of the processor and discard any lower-level structure that interferes with the efficient execution of code.3.1.1RegistersThe interpreter replicates user-visible registers as an array of64-bit quantities in memory. On a64-bit host machine these are native unsigned64-bit integers;on32-bit machines they are two-element struct s.There are several caveats,however.The SPARC archi-tecture supports register windows for integer registers.This was emulated by allocating a large array of64-bit quantities,setting the lowest8to be the global registers,and having a window of24registers slide up and down the array as procedure calls are made and registers are saved and restored.It is important to be able to restore the user stack in order to be able fully to emulate a system call.It is also important to keep SIND’s own stack separate from the user stack,because the interpreter runs in the address space of the user process and so in principle the user process’s stack entries might clobber the interpreter’s stack.Creating and maintaining two separate stacks is discussed below,but the discussion of restoring the user stack is in the Syscall Manager section.Maintaining two separate execution stacks requires a bit of system hacking.The last valid stack frame is left alone,and its stack pointer(pointer to the top of the frame)is savedfor reference.A new page is allocated for the separate stack,and its topmost address is recorded.This topmost address is to become the new frame pointer.Then an explicit save instruction is issued;it creates a new register window,but with the stack pointer pointing into the new page.Then the frame pointer register can be manually set.From that moment on,all further calls should write their stack data to the alternate stack page(s).Apart from protecting the SIND call stack from manipulation by the interpreted program,this also means that SIND’s stack can be mprotect-ed to safeguard its contents when executing code directly on the processor(either issuing traps or when in the fragment cache).Thefloating point registers on the SPARC consist of three overlapping sets of32,64, and128-bitfloating point registers.There are3232-bit,3264-bit registers,and16128-bit registers.The128-bit and64-bit registers overlap completely(e.g.,thefirst128-bit register is the same as thefirst two64-bit registers),and the32-bit registers overlap with the bottom half of the other two.This was implemented as a contiguous region of memory, accessed in different ways depending upon the instruction used(some checking had to be done to make sure no accesses were attempted to non-existent32-bit registers).3.1.2InstructionsAlthough the SPARC v9architecture is64-bit,the instructions are still32-bit,which al-lows backward compatibility(consequently,the software interpreter is also capable of run-ning SPARC v8code).The SPARC has30different instruction formats,grouped together into4major families.However,these formats are all the same length(32bits)and were designed to be quickly parsed by hardware.This permits streamlining the fetch and decode portions of the interpreter.Each instruction format was specified with its own bit-packed struct,and all such structs were grouped together in a union with a normal unsigned32-bit integer.Each format family is distinguished from the others by the two high-order bits ofthe instruction.1Thus the interpreter has jump tables for each instruction format family (actually three jump tables and one explicit function call),that are keyed by the opcode, whose position depends upon the format family.A case statement branches on the two most significant bits to the correct jump table,and then the correct function is called.3.1.3Exceptional ConditionsOccasionally during execution,an instruction will cause an error.The SPARC v9architec-ture manual clearly defines these exceptions,and,for each instruction,specifies which ex-ceptions it can raise.Many of the exceptions are caught by the operating system and used to handle things like page faults and memory errors.Non-recoverable exceptions usually cause the operating system to send a signal to the executing process.To mimic this,if the interpreter decides a given instruction would cause an exception(such as divide-by-zero), then a procedure similar to that used for system calls can be used.The running binary’s state is restored on the stack,and then the interpreter executes the offending instruction directly on the processor.This generates the appropriate operating system action(usually, killing the process).3.1.4Signals and Asynchronous I/OIn the Solaris operating system there are really only two ways of communication between user and supervisor(kernel)code.One,the system call or trap,is discussed in the Syscall Manager section.The other,signals,had to be dealt with differently.Because the SIND system is guaranteed to be loaded before all other libraries,its definitions of functions will take priority(if they’re exported).The SIND interpreter interposes on the signal。

问卷调查设计

问卷调查设计
comments. Leave a space at the end of a questionnaire entitled
"Other Comments." Try to keep your answer spaces in a straight line, either
horizontally or vertically. .
Where did you grow up? __

A. country

B. farm

C. city
5. Produces variability of responses.
What do you think about this report?
A. It's the worst report I've read
What brand of computer do you own? __

A. IBM PC

B. Apple
3. There are two ways to correct this kind of problem.
The first way is to make each response a separate
Use simple and direct language Make it convenient. Always consider the layout of your questionnaire. Leave adequate space for respondents to make
money?
9. Does not use emotionally loaded or vaguely defined words.

Numerical Linear Algebra

Numerical Linear Algebra

letters (and occasionally lower case letters) will denote scalars. RI will denote the set of real
tions to the algorithm, it can be made to work quite well. We understand these algorithmic
transformations most completely in the case of simple algorithms like Cholesky, on simple
LA
Numerical Linear Algebra
Copyright (C) 1991, 1992, 1993, 1994, 1995 by the Computational Science Education Project
This electronic book is copyrighted, and protected by the copyright laws of the United States. This (and all associated documents in the system) must contain the above copyright notice. If this electronic book is used anywhere other than the project's original system, CSEP must be noti ed in writing (email is acceptable) and the copyright notice must remain intact.

GW Invariants and Invariant Quotients

GW Invariants and Invariant Quotients
1991 Mathematics Subject Classification. 14H10, 14L30.
1
2
GW INVARIANTS AND INVARIANT QUOTIENTS
precise meaning of the notations will be given later on): ˆ := exp.dim M g,k (X/ ˆ) = (3 − dim X/ ˆ + k, D /G, A /G)(g − 1) + c1 (X/ /G) · A /G D − dim G := exp . dim .M g,k (X, A)/ = (3 − dim X )(g − 1) + c1 (X ) · A + k − dim G, and therefore ˆ − (D − dim G) = g · dim G. (⋆) D ˆ) is larger From this computation we deduce that in general the space M g,k (X/ /G, A than M g,k (X, A)/ /G, the only exception happening in genus zero. Very shortly, the explanation for this phenomenon is that the projective line is the only one smooth curve which has the property that a topologically trivial, holomorphic principal G-bundle over it is also holomorphically trivial. In higher genera, holomorphic principal G-bundles with fixed topological type depend on ‘moduli’, whose number agrees with the difference (⋆) above. This is the reason why for computing higher ˆ we will need to consider maps into a larger variety X ¯ whose genus invariants of X construction, in the case when G is a torus, is given in lemma 6.1. The article is organized as follows: the first section recalls some basic facts about stable maps and their moduli spaces, the reference being [9]. In section 2 we describe the G-semi-stable points of the moduli space of stable maps M g,k (X, A). The results obtained in this section hold in full generality, no matter what the G-action on X looks like. We obtain the sufficient result (theorem 2.5) which says that a map with image contained in the semi-stable locus of X is G-semi-stable as a point of M g,k (X, A) and a necessary result (corollary 2.4) which says that a stable map representing a G-semi-stable point of M g,k (X, A) does not have its image contained in the unstable locus of X . Section 3 characterizes the semi-stable points of M g,k (X, A) from a symplectic point of view, which will be useful later on in section 6 where we will give an algebro-geometric construction of the space of maps needed for defining certain ‘Hamiltonian invariants’. We compute an explicit formula for the moment map on the space of stable maps which corresponds to a C∗ -action (proposition 3.4), and using it we give (theorem 3.7) a second proof for theorem 2.5. Finally, section 4 closes the first part of the article giving a partial answer to the initial problem, that of comparing the genus zero GW-invariants of X and X/ /G. Theorem 4.1 states, under certain transversality assumptions which are technical in nature, that if G acts freely on the semi-stable locus of X , A is a spherical class for which a representative may be found to lie entirely in the stable locus X ss and ˆ denotes the push-forward class in X/ A /G, then

BTBU_4.2 Audit Partner Specialization and Audit Fees_ Some Evidence from Sweden

BTBU_4.2 Audit Partner Specialization and Audit Fees_ Some Evidence from Sweden

Audit Partner Specialization and Audit Fees:Some Evidencefrom Sweden*MIKKO ZERNI,University of Vaasa1.IntroductionThe purpose of this study is to examine auditor specialization and pricing at the individual partner level.1In the aftermath of major accounting scandals such as ComROAD AG, Enron,Parmalat,Tyco,Waste Management,and WorldCom,regulators and investment communities have been seeking to restore investor confidence in the capital markets.The response worldwide has been increases in regulation,and in these reforms accounting and auditing have been identified as priority areas to‘‘fix’’.For instance,with the aim of increasing audit market transparency,the amended European Union’s(EU’s)8th Directive requires the disclosure of engagement partner identity.2Currently,the Public Company Accounting Oversight Board(PCAOB)in the United States is considering a similar requirement.On October6,2008,the U.S.Treasury’s Advisory Committee on the Audit-ing Profession(ACAP)issued itsfinal report,which recommends,among other things,‘‘urging the PCAOB to undertake a standard-setting initiative to consider mandating the engagement partner’s signature on the auditor’s report’’(ACAP Report,October6,2008, at VII:19).3According to the ACAP’s recommendation,the requirement for the engage-ment partner to sign the audit report could improve audit quality in two ways:‘‘First,it might increase the engagement partner’s sense of accountability tofinancial statement *Accepted by Ferdinand A.Gul.An earlier draft of this paper was entitled‘‘Audit Partner Specialization, Audit Fees,and Auditor-Client Alignments’’.I appreciate the comments received from Ferdinand A.Gul (the associate editor),two anonymous reviewers,Pekka Alatalo(KPMG Finland),Jean C.Bedard(discus-sant),Andy Conlin,Ann Gaeremynck,Kaarina Halonen(PWC Finland),Seppo Ika heimo,Henry Jarva, Juha Joenva a ra,Juha-Pekka Kallunki,Eija Kangas(KPMG Finland),Robert Knechel,Anna-Maija Lantto, Christophe Van Linden,Lasse Niemi,Henrik Nilsson,Mervi Niskanen,Jukka Perttunen,Peter Pope, Markku Rahiala,Veijo Riistama(PWC Finland),Petri Sahlstro m,Stefan Sundgren,Risto Tuppurainen, Sofie Vandenbogaerde,Ann Vanstraelen,Markku Vieru,Marleen Willekens,and Erik A stro m(Ernst& Young Sweden).I would also like to thank the participants at the24th Contemporary Accounting Research Conference in Montreal Canada(2009),AFI seminar at Katholieke Universiteit Leuven(2011)and the AFAR workshop in Vaasa(2008)for their comments.I wish to thank Tuomas Anttila,Marja Kauppinen, and Harri Lempola for their excellent research assistance.Financial support received from the NASDAQ OMX Nordic Foundation,the Finnish Foundation for the Advancement of Securities Markets,the Founda-tion for Economic Education,the Ostrobothnia Cultural Foundation,and the Finnish Cultural Foundation is gratefully acknowledged.This research is part of research projects by the Academy of Finland(Grant Numbers140000and126630).All remaining errors are mine alone.1.The terms‘‘auditor specialization’’and‘‘auditor expertise’’,as well as the terms‘‘engagement partner’’,‘‘auditor’’,and‘‘audit partner in charge’’are used interchangeably in this study.2.The new EU directive obligesfirms to disclose the identities of individual auditor(s)responsible for theengagement.Specifically,Article28of the directive2006⁄43⁄EC states:‘‘Where an auditfirm carries out the statutory audit,the audit report shall be signed by at least the statutory auditor(s)carrying out the statutory audit on behalf of the auditfirm.’’3.The comment period on the concept release ended on September11,2009,and according to the PCAOB’sOffice of the Chief Auditor’s standard-setting agenda,‘‘the Board’s consideration of next steps is pending further action’’(PCAOB,October2010,p.7).Available at:/News/Events/Documents/ 10132010_SAGMeeting/OCA_standards-setting_agenda.pdf.Contemporary Accounting Research Vol.29No.1(Spring2012)pp.312–340ÓCAAAdoi:10.1111/j.1911-3846.2011.01098.xAudit Partner Specialization and Audit Fees313 users,which could lead him or her to exercise greater care in performing the audit. Second,it would increase transparency about who is responsible for performing the audit, which could provide useful information to investors and,in turn,provide an additional incentive tofirms to improve the quality of all of their engagement partners.’’Some have compared the initiative to thefinancial statement certification requirement by top manage-ment stipulated in Section302of the Sarbanes-Oxley Act,arguing that it should help focus engagement partners on their existing responsibilities(see,e.g.,Carcello,Bedard, and Hermanson2009:79).4Changes in legislation and initiatives requiring the disclosure of the identities of individ-ual auditors carrying out audits implicitly acknowledge that a public company audit involves a substantial amount of work by highly skilled individual practitioners exercising their own professional judgment.It is the lead engagement partners working in the city level audit offices who play a central role in planning and implementing the audit and ulti-mately in determining the appropriate type of audit report to be issued to the client(e.g., Ferguson,Francis,and Stokes2003).Consequently,the engagement partner may play an essential role in the(perceived)audit quality beyond auditfirm size and(industry)special-ization at the national and office level and should hence not be ignored.The level of audit effort and fees depends on the client’s agency-driven demand for external auditing and on supply-side factors,such as auditfirm size,auditor expertise and auditor-perceived risk factors(e.g.,DeFond1992;O’Keefe,Simunic,and Stein1994; Gul and Tsui1998;Bell,Landsman,and Shackelford2001;Johnstone and Bedard2001, 2003;Gul and Goodwin2010;Causholli,De Martinis,Hay,and Knechel2011).From the supply-side perspective,auditors are expected to respond to the higher probability of any irregularities or accounting misstatements by increasing audit effort and charging higher fees.For instance,Gul and Tsui(1998)adopted a supply-side perspective and found that higher inherent risks associated with free cashflows are associated with higher audit effort and higher consequent fees.From the demand-side perspective,the appointment of a higher-quality auditor can serve as a signal of an enhanced quality of financial disclosure,which will potentially lead to greater value for the audit client by reducing some agency costs.Firm insiders,especially those in the heart of monitoring function(e.g.,independent directors on the boards and audit committees),may be willing to increase audit coverage to create a positive perception about thefinancial reporting quality.The positive perception will possibly facilitate thefirm to attract investments and fund profitable projects and allowfirm insiders to protect their reputation capital,avoid legal liability,and promote shareholder interests.Consistent with the demand-side per-spective,several studies report evidence suggesting that outside directors who act dili-gently demand high quality audits and pay higher audit fees(e.g.,Carcello,Hermansson, Neal,and Riley2002;Abbott,Parker,Peters,and Raghunandan2003;Knechel and Willekens2006).Auditing is generally viewed as a differentiated service with substantial variation observed in audit(effort)fees,even after controlling for observable factors such asfirm size and complexity.This differentiation allows clients some choice over the level of audit scrutiny even within audits conducted by the same(tier)auditfirms.An important means 4.Recent empirical evidence supports the view that the Sarbanes-Oxley Act Section302chief executive offi-cer(CEO)and chieffinancial officer(CFO)certification requirement had a positive effect onfinancial reporting quality.For instance,Cohen,Krishnamoorthy,and Wright(2010)report that68percent of practicing auditors interviewed believe that the certification requirement has had a positive effect on the integrity offinancial reports.Moreover,to the extent that the disclosure of engagement partner identity will increase accountability,it may thereby improve audit decisions and judgments.In particular,some studies in the auditing context report evidence indicating that accountability reduces information process-ing biases,and increases consensus and self-insight(e.g.,Johnson and Kaplan1991;Kennedy1993).CAR Vol.29No.1(Spring2012)314Contemporary Accounting Researchof audit product differentiation is through investments in specialization(Simunic and Stein1987;Liu and Simunic2005).By specializing in certain industries,certain size groups,or companies with similar risk profiles,individual auditors may be able to differ-entiate their product from those of nonspecialist audit partners(Simunic and Stein1987; Liu and Simunic2005).The Swedish Code for Corporate Governance,for instance, implicitly recognizes auditor specialization in large public companies as one relevant piece of information when assessing the quality of external auditing.5Specifically,the Code rec-ommends that information on‘‘the audit services performed by the auditor or the auditor in charge in other large companies...and other information that may be important to shareholders in assessing the competence and independence of the auditor,or auditor in charge must be disclosed in a corporate governance report on company’s homepage’’(Ori-ginal Code Sections2.3.2and2.3.3).To justify their presence in the audit market and the potential fee premiums attached to their services,the client must perceive some benefit in hiring a specialist auditor.The premium may,for instance,be attributed to the signal value of hiring a specialist auditor or to the superior advisory or other services provided by that auditor(Titman and Trueman1986).Given that a demand exists for special-ist auditors,that demand gives those auditors a greater‘‘power’’in pricing relative to nonspecialists.This paper is motivated by the lack of archival research examining issues related to engagement partner specialization.In the present study,the use of Swedish data makes it possible to construct individual audit partner client portfolios because each audit report discloses the name of the audit engagement partner.Thus,information on the identity of the audit partner in charge is observable to users offinancial statements and thereby potentially affects market-assessed perception of ex ante audit rma-tion on the sizes and compositions of the Big4audit partner-specific client portfolios provides an interesting opportunity to contribute to a more thorough understanding of auditor specialization.More specifically,it is possible to examine whether the perceived audit quality is affected not only by the brand name of thefirm,but also by the charac-teristics and reputation of the engagement partner.In essence,if auditing expertise were wholly transferable and therefore uniformly distributed across audit partners within the firm,any clientfirm would be indifferent to having any audit partner within the Big4 auditors(within a particular Big4auditfirm⁄office)to conduct the audit.Additionally, there would be no a priori reason why clients would be willing to pay any audit partner related premiums.The empiricalfindings indicate systematic differences between audit partner clienteles, suggesting audit partner specialization in different industries and in different size groups. Thisfinding is consistent with Liu and Simunic2005,who argue that auditor specializa-tion could be a competitive response by either an auditfirm or an individual audit part-ner to induce efficient audits for different types of clients,thereby gaining a limited monopoly power(and earning rents)over the clients in which they specialize.Further-more,consistent with the view that there are returns on investing in specialization,analy-ses of audit fees indicate that both audit partner industry specialization and specialization in large public companies are recognized and valued byfinancial statement users and⁄or by corporate insiders,resulting in higher fees within these engagements.According to the empirical analyses,the highest fees are earned by engagement partners who are both 5.The Swedish Corporate Governance Board is responsible for promoting and developing the Code.On July1,2005,the Stockholm Stock Exchange began applying the Swedish Code of Corporate Governance.The Code applies to all Swedish companies listed at the Stockholm Stock Exchange and foreign companies that are listed at the same exchange and whose market capitalization exceeds SEK3billion.For more information,see http://www.corporategovernanceboard.se/.CAR Vol.29No.1(Spring2012)Audit Partner Specialization and Audit Fees315 industry and publicfirm specialists.The results may be interpreted to mean that the appointment of a specialist engagement partner is associated with higher(perceived)audit quality,thus justifying the fee premium.6Collectively,thefindings of this study support the view that clientfirms infer audit quality at least to some extent from the characteris-tics of the individual audit partner in charge.The remainder of the paper is organized as follows.Section2reviews the relevant lit-erature and describes some relevant features of the Swedish audit market.Section3pre-sents the hypothesis,and section4describes the data.Section5describes the methodology used,section6presents the empirical results,and section7concludes the study.2.Literature reviewNational versus local view of the auditor–client relationshipPrior audit research literature is dominated byfirm-wide analyses treating the whole accountingfirm as the focal point and investigating whether and how auditfirm char-acteristics,such as size and industry specialization at the nationalfirm level,affect the auditor–client relationship(e.g.,Simunic and Stein1987;Francis and Wilson1988;Bec-ker,DeFond,Jiambalvo,and Subramanyam1998;Francis and Krishnan1999).All these studies implicitly assume that through standardizedfirm-wide policies and knowl-edge sharing(e.g.,through training materials,industry-specific databases,internal benchmarks for best practices,audit system programs,and internal consultative prac-tices),all audits across practice offices and audit partners within an auditfirm are uniform.7However,in practice,it is the individual audit partners from city-level practice offices who are creating and taking care of the relationship,contracting with the client,adminis-tering the audit engagement,directing the audit effort,interpreting the audit evidence,and finally issuing the appropriate audit report(Ferguson et al.2003).Because of this decen-tralized organizational structure and because individual audit partners and their character-istics vary across engagements,a research approach allowing each individual partner to be a unique and relevant unit of analysis might be a more reasonable approach than assum-ing that all audits within an auditfirm are uniform.Consistent with this intuition,a growing number of recent audit studies have changed the direction from afirm-wide to an office-level view of auditfirms(e.g.,Reynolds and Francis2000;Ferguson et al.2003;Francis,Reichelt,and Wang2005;Francis and Yu 2009;Reichelt and Wang2010;Choi,Kim,Kim,and Zang,2010).The local stream of audit literature acknowledges the likelihood that part of an auditor’s expertise is uniquely held by individual professionals through their personal knowledge of clients and cannot be 6.From the risk-based audit supply view,an alternative explanation for the observed specialist audit partnerfee premium is that the clients of specialist audit partners are systematically riskier than the clients of non-specialists in dimensions other than those already controlled for in the empirical fee model(or in thefirst-stage selection model of the Heckman two-stage approach).Despite using a two-stage Heckman1978pro-cedure and including several audit risk-related control variables in the audit fee model,the empiricalfind-ings remain susceptible to the concern that some of the omitted risk factors recognized and priced by the specialist audit partners that simultaneously determine the alignment of audit partners with engagements would explain the outcome.7.Examples of the information technology systems used in knowledge sharing include KPMG’s KWorld TM,PriceWaterhouseCoopers’s TeamAsset TM and KnowledgeCurve TM,and Ernst&Young’s Knowledge-Web TM.See Vera-Munoz,Ho,and Chow2006for factors affecting knowledge sharing within interna-tional accountingfirms,and Banker,Chang,and Kao2002for a detailed description of an international accountingfirm’s implementation of audit software and groupware for knowledge sharing.CAR Vol.29No.1(Spring2012)316Contemporary Accounting Researchreadily captured and distributed by thefirm to other offices and clients(e.g.,Ferguson et al.2003).8The results of the empirical studies adopting a local perspective tend to provide a bet-ter understanding of the operations of a Big4auditfirm than do studies taking a national perspective.For instance,there is evidence that auditors’reputation for industry expertise is neither strictly national nor strictly local in character.Auditors who are both city and industry leaders are reported to earn fee premiums both in Australia and in the United States,suggesting that there is both a national and local office reputation effect in the pric-ing of industry expertise(Ferguson et al.2003;Francis et al.2005).Moreover,both these studiesfind thatfirm-level industry specialists alone do not earn statistically significant premiums.In another study,Reichelt and Wang(2010)use U.S.data and three proxies for audit quality(abnormal accruals,clientfirms’likelihood of meeting or beating ana-lysts’earnings forecasts by one penny per share,and the propensity to issue a going con-cern audit opinion)andfind evidence consistent with the view that audit quality is higher when the auditor is both a national and city-specific industry expert.Recent and concurrent studies have pushed the local analysis still one step further to the engagement partner level.A growing number of studies use engagement partner data (mainly from Australia and Taiwan)and examine issues such as the relationship between engagement partner tenure and audit quality,producing mixedfindings(e.g.,Carey and Simnett2006;Chen,Lin,and Lin2008;Chi,Huang,Liao,and Xie2009).In a recent study,Chin and Chi(2009)use a sample of listedfirms in Taiwan to investigate the associ-ation between auditor industry expertise and restatement likelihood at the partner level and at the auditfirm level simultaneously.Their evidence suggests that the differences in restatement likelihood due to industry expertise is mainly attributable to the partner-level experts rather than to thefirm-level experts.In summary,the recent empirical evidence suggests that auditor expertise has a strong local dimension.The central conceptual question in all the national versus local studies relates to the degree to which there is a transfer of expertise from office-based accounting professionals to other auditors and offices within thefirm(Ferguson et al.2003;Francis2004).There are several factors that may deter the transfer of expertise within organizations,including audit firms(see,e.g.,Szulanski1994,2000;Nonaka and Takeuchi1995).With respect to audit firms,Vera-Munoz et al.(2006)enumerate several reasons why it is difficult for partners to share knowledge with other partners.First,a considerable amount of knowledge in audit firms can be difficult to document or transfer.9Second,even if an auditfirm manages to 8.As noted by Francis2004,another reason for this development is that Big4market shares continue toexpand globally,leading to low power in research designs of studies comparing large and small auditors because there is such low variance in the experimental variable(i.e.,most observations are audited by large Big4auditors).For instance,according to a recent Government Accountability Office(GAO) report,in2006the largest fourfirms collected94percent of all audit fees paid by public companies.More-over,according to the same report,82percent of the Fortune1000companies saw their choice of auditors as limited to three or fewerfirms,and about60percent viewed competition in their audit market as insuf-ficient(GAO2008).9.According to Polanyi1966,knowledge comes in two types:explicit and tacit.The former is amenable tocodification,while the latter is anchored in individual personal beliefs,experiences,and values.For exam-ple,knowledge of generally accepted accounting principles(GAAP)with regard to fair value requirements is explicit knowledge,while an auditor’s insights as to how a client’s management develops fair value esti-mates and whether those estimates conform to GAAP represents tacit knowledge(Vera-Munoz et al.2006).Tacit knowledge is subconsciously understood and applied and is therefore not easily articulated (Polanyi1966).Prior studies further show that most of the knowledge in any organization,including an auditfirm,is tacit knowledge(Bonner2000;Knechel2000).Consequently,because differences in knowl-edge between individual auditors within afirm lie primarily in their respective tacit knowledge,which is not easily transferred,it is unlikely thatfirm-level practices completely smooth out the differences in the levels of individual audit partner expertise.CAR Vol.29No.1(Spring2012)Audit Partner Specialization and Audit Fees317 collect extensive databases and otherfirm-wide knowledge,individual auditors still need to use their own judgment in selecting and applying relevant pieces of information given the task at hand.Third,knowledge-sharing through information technology–based expert knowledge systems is not automatically embraced by everyone.Finally,evaluation appre-hension,performance-based compensation schemes and individual auditors’pursuit of per-sonal benefits and power may deter auditors from sharing what they know.In essence, holding on to information that other peers do not have may ceteris paribus offer competi-tive advantage against those peers,when auditors,like other rational players in the econ-omy,attempt to maximize their own(economic)interests.Collectively,factors identified above may partly explain the tendency of local audit studies to provide a better understand-ing of the auditor-client relationship thanfirm-wide analyses.Factors affecting expertiseThe Merriam-Webster dictionary defines an‘‘expert’’as having,involving,or displaying a special skill or knowledge derived from training or experience.The psychological literature on expertise has reported two importantfindings relevant to the present study.First, domain-specific knowledge is the essential determinant of expertise.Second,expert knowl-edge is gained through many years of on-the-job experience(e.g.,Chi,Glaser,and Rees 1982;Glaser and Chi1988;Glaser and Bassok1989;Lapre,Mukkerjee,and Van Was-senhove2000).In other words,intensive practice and the repetition of similar tasks are required to build expertise.When applied to auditing,thesefindings suggest that having serviced many similar clients in the past may help auditors to develop a specialized knowl-edge of what these clients do and the challenges and issues they face,thereby creating in-depth knowledge of specific types of clients,leading to higher-quality audits and higher fees in these engagements.Consequently,in the present study,the terms‘‘auditor special-ization’’and‘‘auditor expertise’’are used to refer to the extent of auditors’prior audit experience with similar clients,for instance,clientfirms belonging to the same industry or size group.In other words,it is assumed that specialization is needed to gain deep exper-tise.By specializing in certain industries,certain size groups,or companies with similar risk profiles,individual auditors may be able to develop and supply the differentiated ser-vice that clients demand and that competitorsfind difficult to duplicate(Simunic and Stein 1987;Liu and Simunic2005).Moreover,auditors will only develop a specialist reputation if this increases the credibility offinancial reporting and attracts clients.Auditor specialization and audit qualityFor a long time,standard-setters,quasi-regulatory bodies,and empirical audit research have suggested that differences in the level of auditor industry expertise may be one source of variation in audit quality(Hogan and Jeter1999;Gramling and Stone2001;GAO 2008).This body of literature assumes that audit issues are nested within an industry and that accountingfirms or individual auditors with many clients within a particular industry have more opportunities to acquire the kind of profound industry knowledge that leads to industry expertise.It is also argued that the heavy investments of industry specialist audi-tors in technologies,physical facilities,personnel,and organizational control systems pro-vide them with both incentives and abilities that make them more likely than nonspecialist auditors to detect and report any irregularities or misrepresentations in clientfirms’accounts(Simunic and Stein1987).Prior archival studies tend to document a positive association between auditors’indus-try expertise and the quality offinancial reporting(e.g.,Carcello and Nagy2002,2004; Balsam,Krishnan,and Yang2003;Krishnan2003,2005;Gul,Fung,and Jaggi2009).Spe-cifically,industry specialist auditors have been found to be less likely to be associated with Securities and Exchange Commission enforcement actions(Carcello and Nagy,2004).CAR Vol.29No.1(Spring2012)318Contemporary Accounting ResearchTheir clients are also reported to have a lower probability offinancial fraud(Carcello and Nagy2004,2004),smaller amounts of abnormal accruals,and higher earnings response coefficients(Balsam et al.2003;Krishnan2003,2005).In addition,a recent study by Gul et al.2009reports evidence suggesting that auditor industry specialization is likely to reduce the association between shorter auditor tenure and lower earnings quality.Behavioral research using experimental approaches has examined industry specializa-tion at the individual auditor level.The results of these studies suggest that an individual auditor’s expertise is tied not only to each individual professional and his or her deep per-sonal knowledge of clients but also to the innate abilities of each individual(e.g.,Bonner and Lewis1990;Libby and Tan1994;Owhoso,Messier,and Lynch2002).Overall,the results of studies on auditor specialization suggest that industry specialist auditors deliver higher-quality audits than do nonspecialists and that this difference in quality is also rec-ognized by the audit market.Even though the auditor specialization literature has focused almost entirely on indus-try specialization,it is plausible that there are also other types of auditor specialization besides industry specialization.Accordingly,an investigation of the(Swedish)homepages of the Big4firms reveals that each of thefirms has structured their national practices along both industry lines and client size.All have at least two business lines based on client size:small and medium-sized companies and large public⁄international companies. All fourfirms also market a wider variety of specialized expertise,such as specialization in owner-manager companies,family businesses,public sector organizations,and nonprofit organizations.Moreover,as noted already,the Swedish Code for Corporate Governance implicitly recognizes auditor specialization in working with different size groups by recom-mending the disclosure of‘‘the audit services performed by the auditor or the auditor in charge in other large companies’’,viewing this as a relevant piece of information for users offinancial statements.Performing audits of large,complex high-profile clients is likely to require auditor expertise widely different from that required for audits of smaller and simpler closely held clients.A public listing is an important part of auditor business risk(Johnstone and Bedard2003).Because companies listed on a stock exchange receive more media attention and therefore pose a higher litigation and reputational risk,they may require a specialist auditor(Johnstone and Bedard2003).Another clear distinction between audits of public and private companies relate tofinancial reporting standards.10In Swe-den,as in all EU member countries,publicly listed companies are required to follow IFRS standards in theirfinancial reporting.Whereas Swedish private companies are also allowed to follow IFRS standards in their consolidatedfinancial statements,they tend to follow national standards(Bokfo ringsna mndens Anvisningar),which include several significant simplifications compared to IFRS.Accordingly,in the present study,com-pany size and listing status in particular is used as a proxy for client complexity and risk profile.The aforementioned differences between audits of public and private companies make it more likely that client companies will value auditor specialization in particular size groups(i.e.,public versus privatefirms).They may also suggest that specialization in a certain size group will overlap somewhat with industry specialization.It appears intuitively more valuable for publicly listed client companies seeking a higher level of audit assurance to hire an engagement partner with relevant industry experience on other large companies with similarfinancial reporting requirements,rather than hire an engagement partner with industry experience only on smaller private companies.This potential overlap is further addressed in the empirical results section below.10.I wish to thank an anonymous reviewer for making this point.CAR Vol.29No.1(Spring2012)。

Improved Iteratively Reweighted Least Squares for Unconstrained Smoothed Lq Minimization

Improved Iteratively Reweighted Least Squares for Unconstrained Smoothed Lq Minimization
2
+ |xi |2 )q/2−1 .
(k)
Department of Mathematics, The University of Georgia, Athens, GA 30602. Dept. of Applied and Computational Mathematics, Rice University, Houston, TX. This author is partly supported by ARL and ARO grant W911NF-09-1-0383 and AFOSR FA9550-10-C-0108. ‡ wotao.yin@. Dept. of Applied and Computational Mathematics, Rice University, Houston, TX. This author is partly supported by NSF grant DMS-0748839 and ONR Grant N00014-08-1-1101.
X ∗,
subject to PΩ (X ) = PΩ (M ),
(1.1)
where [m] := {1, 2, . . . , m}, the nuclear norm X ∗ is the sum of the singular values σi (X ) of X , i.e. r X ∗ = i=1 σi (X ), and PΩ (X ) = PΩ (M ) is short for Xij = Mij , (i, j ) ∈ Ω. The work [29] studies the lowrank matrix recovery problem with constraint A(X ) = A(M ) for general linear operator A : Rm×n → Rp . Various types of algorithms have been proposed for solving problem (1.1), and many of them are extensions or adaptations of their predecessors for sparse vector recovery. They include, but not limited to, the singular value thresholding (SVT) algorithm [3] based on the linearized Bregman algorithm [36, 28], fixedpoint continuation code FPCA [24] extending FPC [18], the code APGL [32] extending [2], and the code [35] based on the alternating direction method [16]. Algorithms with no vector-recovery predecessors include OptSpace [20] and LMaFit [34], which are based on explicit factorizations M = U SV ∗ and M = XY and nonlinear least-squares formulations minU,S,V PΩ (U SV ∗ − M ) F and minX,Y PΩ (XY − M ) F , respectively. q Besides the above, the nonconvex q quasi-norm x q q = i |xi | , 0 < q < 1, and its variants have been used to develop algorithms for recovering sparse vectors in [7, 9, 10] and low-rank matrices in [26, 13]. First of all, compared to 1 norm x 1 , x q q for 0 < q < 1 makes a closer approximation to the “counting norm” x 0 , which is the number of nonzero entries of x. It is shown in [8] that assuming certain restricted isometry properties (RIPs) of the sensing matrix A, a sparse vector xo ∈ RN is the q minimizer of Ax = b, where b := Axo can have fewer observations than needed by convex 1 minimization. Works [15, 31] derive sufficient conditions in terms of RIP of A for q minimization to recover sparse vectors that are weaker than those known for 1 minimization. However, the q quasi-norm is nonconvex for q < 1, and q minimization is generally NP-hard [17]. Instead of directly minimizing the q quasi-norm, which most likely ends up with one of its many local minimizers, algorithms [7, 9, 10] solve a sequence of smoothed subproblems. Specifically, [7] solves reweighted (k) , the algorithm generates a new iterate x(k+1) by minimizing 1 subproblems: given an existing iterate x (k) q −1 . To see how it relates to q quasi-norm, one can let x(k) = x, i wi |xi | with weights wi := ( + |xi |) = 0, and 0/0 be 0 and then get i wi |xi | = i |xi |q = x q q . On the other hand, [9, 10] solves reweighted q 2 2 (more precisely, least-squares) subproblems: at each iteration, they approximate x q by i wi |xi | with weights wi := (

汽车后桥总成设计解放轻卡

汽车后桥总成设计解放轻卡

摘要按照车桥可否传递驱动力,汽车车桥分为驱动桥和从动桥。

驱动桥的结构型式按齐整体布置来讲共有三种,即普通的非断开式驱动桥,带有摆动半轴的非断开式驱动桥和断开式驱动桥。

本设计对象是轻型低速载货汽车的后驱动桥。

本设计完成了轻型低速载货汽车的后驱动桥中主减速器、差速器、减震器、钢板弹簧及桥壳等部件的设计。

按照轻型低速载货汽车的后驱动桥的要求,通过选型,肯定了主减速器传动副类型,差速器类型,驱动桥半轴支承类型减震器类型和钢板弹簧类型。

通过设计计算,肯定了主减速比,主、从动锥齿轮、差速器、半轴、减震器、钢板弹簧和桥壳的主要参数和结构尺寸。

利用Pro/E软件画出所有零部件的三维视图及装配图和总装配图然后生成工程图,通过主要零部件的校核计算和利用CAD对主要零部件就行二维画图,肯定所设计的能够知足设计要求。

关键词:汽车后桥;主减速器;差速器;减震器;钢板弹簧AbstractAccording to the axle can transfer the driving force, the car axle is divided into a drive axle and a driven axle. Drive bridge structure according to the general layout, with a total of three species, namely ordinary non-break drive bridge, a swing axle non-break drive axle and a broken axle. The object of this design is light-duty low-speed truck drive axle.Completion of the design of light truck speed rear driving axle main reducer, differential, shock absorber, a leaf spring and the axle housing and other components of the design. In this paper, according to the light of low-speed truck drive axle requirements, through the selection, determination of main reducer transmission pair type, differential type, drive axle bearing type shock absorber type and the leaf spring type. Through design calculation, determine the main reduction ratio, main, the driven bevel gear, differential gear, axle, shock absorber, steel plate spring and axle housing main parameters and dimensions.Using Pro/E software to draw all the parts of the three-dimensional- view and assembly drawings and assembly drawings and then generate engineer- ing drawing, the main components of the calculation and use of CAD on the key parts on the line drawing, determine the design can meet the design requirements.Key Words: automobile rear axle ;main reducer;differential device ;shockabsorber; plate spring目录摘要 (Ⅰ)Abstract (Ⅱ)第1章绪论 (1)本课题的来源、大体前提条件和技术要求 (1)本课题要解决的主要问题和设计整体思路 (1)预期的功效 (2)国内外研究现状及发展趋势 (2)课题研究内容 (3)第2章汽车主参数的整体设计 (4)设计参数与设计目标 (4)汽车轴数及驱动形式的选择 (4)轴数 (4)驱动形式 (4)轻型载货汽车质量参数选择 (4)整车装备质量 (5)汽车的总质量 (5)汽车轴荷分派 (5)汽车轴距、后轮距及悬架长度设计 (6)轴距 (6)后轮距 (7)汽车后悬架长度 (8)第3章后桥主要零部件的设计计算 (9)悬架的的设计计算 (9)悬架的的结构形式分类 (9)悬架主要参数的肯定 (10)影响平顺性的参数 (10)影响操纵稳定性的参数 (11)钢板弹簧的设计计算 (11)钢板弹簧的布置方案 (11)钢板弹簧主要参数肯定 (11)减震器的设计计算 (19)减震器类型 (19)减震器的结构和工作原理 (19)减震器的结构设计及计算 (20)相对阻尼系数的肯定 (20)减振器阻尼系数的肯定 (21)最大卸荷力的肯定 (21)减振器工作缸直径的肯定 (22)工作缸壁厚的计算与校核 (23)活塞杆与活塞的设计 (24)活塞尺寸的计算 (24)底阀的设计 (25)减震器装配进程的三维视图 (27)差速器的设计计算 (30)差速器的结构形式的选择 (30)差速器齿轮的大体参数选择 (31)行星齿轮数量的选择 (31)行星齿轮球面半径的肯定 (31)行星齿轮和半轴齿轮齿数的计算 (31)行星齿轮和半轴齿轮的节锥角及模数的计算 (32)压力角的肯定 (32)行星齿轮轴直径及支承长度 (32)差速器直齿锥齿轮的强度校核 (35)主减速器的设计计算 (37)主减速比的肯定 (37)主减速齿轮计算载荷的计算 (38)主减速齿轮大体参数的选择 (39)第4章汽车后桥其它零部件的设计及后桥总装 (42)汽车驱动桥的设计 (42)汽车驱动桥盖的设计 (43)汽车差速器壳的设计 (44)汽车差速器轴承的选用 (44)汽车差速器轴承座的选型设计 (45)汽车半轴的选型设计 (45)U 型螺栓设计 (46)汽车后桥总装 (46)差速器与主减速器的装配 (46)后桥总装配 (48)后桥总装配剖视图 (51)结论 (52)致谢 (53)参考文献 (54)CONTENTSAbstract (Ⅰ)Contents (Ⅲ)Chapter1 Introduction (1)Topic source basic premise and technical requirement (1)This topic to solve the main problems of the design (1)The expected results (2)The domestic research situation and development trend (2)Subject research contents (3)Chapter 2 Car Lord of the overall design parameters (4)Design parameters and design targe (4)Car and driving shaft for the choice of the form (4)Shaft severa (4)Drive form (4)Light parts of autom obile quality parameter selection (4)Vehicle equipment quality (5)The total quality car (5)Car shaft charge distribution (5)Car wheelbase after the length design (6)Wheelbase (6)reartread (7)Automobile rear suspension length (8)Chapter 3 Major parts of the rear axle design calculation (9)Suspension design calculations (9)Suspension structure of the classification (9)Suspension of the main parameters of the set (10)Influence of the parameters of the smooth (10)Influences of the parameters of the steering stability (11)Leaf spring design calculations (11)Leaf spring arranging schemes (11)Steel spring main parameters (11)Shock absorber design calculation (19)Track of shock absorber type (19)Shock absorber structure and work principle (19)Shock absorber and structure design of calculation (20)Track to determine the relative damping coefficient (20)Shock absorber damping coefficient determinations (21)Biggest unloading the determination of force (21)Shock absorber work to determine the diameter (22)Work cylinder wall thickness calculation and checking (23)Piston rod and the piston design (24)Piston size calculation (24)Bottom valve of design (25)Shock absorber view of the assembly process (27)Differential design calculation (30)The choice of the form of the structure of the differentia (30)The differential gears basic parameter selection (31)Planetary gear number of the choice (31)Planetary gear sphere to determine the radius (31)Planetary gear and half shaft gear gear calculation (31)Planetary gear and half shaft section of gear (32)Pressure Angle sure (32)Planetary gear shaft diameter and length of supports (32)Spur bevel gear differential of intensity (35)The Lord the design of the speed reducer is calculated (37)The determination of the slowdown (37)Lord the reduction gear of the calculation (38)Lord the reduction gear basic parameters selection (39)Chapter 4 Cars driving axle other parts design (42)The design of the car drive axle (42)The design of the car drive axle of cover (43)The design of the car differential shells (44)The selection of car differential bearing (44)Car differential of the bearing type design (45)Car half shaft of the selection of the design (45)U bolt design (46)Car driving axle assembly (46)Differential and the assembly of the Lord reducer (46)Driving axle final assembly (48)Driving axle always assembly section (51)Conclusion (52)Thanks (53)References (54)第1章绪论本课题的来源、大体前提条件和技术要求a. 本课题的来源:轻型载货汽车在汽车生产中占有大的比重。

高三英语计算机语言单选题40题

高三英语计算机语言单选题40题

高三英语计算机语言单选题40题1. When you are programming, you often need to use a(n) ______ to store data.A. algorithmB. variableC. functionD. loop答案:B。

本题考查计算机语言中的常见词汇。

选项A“algorithm”意为“算法”;选项B“variable”指“变量”,在编程中用于存储数据,符合题意;选项C“function”是“函数”;选项D“loop”是“循环”。

2. In computer programming, a(n) ______ is a set of instructions that tells the computer what to do.A. codeB. scriptC. commandD. syntax答案:A。

选项A“code”指“代码”,是一组指令;选项B“script”通常指“脚本”;选项C“command”意为“命令”;选项D“syntax”表示“语法”。

本题强调的是一组指令,所以选A。

3. Which of the following is NOT a type of programming language?A. PythonB. ExcelC. JavaD. C++答案:B。

选项A“Python”、选项C“Java”和选项D“C++”都是常见的编程语言;选项B“Excel”是电子表格软件,不是编程语言。

4. The process of finding and fixing errors in a program is called ______.A. debuggingB. compilingC. optimizingD. documenting答案:A。

“debugging”意为“调试”,即查找和修复程序中的错误;“compiling”是“编译”;“optimizing”指“优化”;“documenting”表示“文档化”。

Li Coefficients for Automorphic L-Functions

Li Coefficients for Automorphic L-Functions

ρ∈Z
If the multiset Z omits the value ρ = 1 then the sums ℜ(λn (Z )) := 1 ℜ 1 − (1 − )n ρ (1.7)
ρ∈Z
converge absolutely for all nonpositive integers n ≤ 0. The positivity condition ℜ(λn (Z )) ≥ 0 for n ≤ 0 then implies that all ρ lie in the half-plane ℜ(s) ≤ 1 2 . If the multiset Z omits the value ρ = 0, then the sum (1.7) then converges absolutely for n ≥ 0, and the positivity condition ℜ(λn (Z )) ≥ 0 for n ≥ 0 implies that all ρ lie in the half-plane ℜ(s) ≥ 1 2 . Combining these criteria, for multisets Z that omit both 0 and 1 the positivity condition ℜ(λn (Z )) ≥ 0 for all integers n implies that all ℜ(ρ) = 1 2 . If the multiset Z is also invariant under the symmetry ρ → 1 − ρ ¯, so that ℜ(λn (Z )) = ℜ(λ−n (Z )), it suffices to check this positivity 1 . Finally, if Z omits the values condition ℜ(λn (Z )) ≥ 0 for n > 0 to conclude that all ℜ(ρ) = 2 1 0 and 1 and the sum ρ∈Z ρ is ∗-convergent, then the coefficients λn (Z ) are well-defined for all integers n by the following ∗-convergent sum: λn (Z ) :=

astm材料与实验标准[1].e112-2004

astm材料与实验标准[1].e112-2004

Designation:E112–96(Reapproved2004)e1Standard Test Methods forDetermining Average Grain Size1This standard is issued under thefixed designation E112;the number immediately following the designation indicates the year of original adoption or,in the case of revision,the year of last revision.A number in parentheses indicates the year of last reapproval.A superscript epsilon(e)indicates an editorial change since the last revision or reapproval.This standard has been approved for use by agencies of the Department of Defense.e1N OTE—Reference(2)was editorially corrected in May2006.INTRODUCTIONThese test methods of determination of average grain size in metallic materials are primarily measuring procedures and,because of their purely geometric basis,are independent of the metal or alloy concerned.In fact,the basic procedures may also be used for the estimation of average grain, crystal,or cell size in nonmetallic materials.The comparison method may be used if the structure of the material approaches the appearance of one of the standard comparison charts.The intercept and planimetric methods are always applicable for determining average grain size.However,the comparison charts cannot be used for measurement of individual grains.1.Scope1.1These test methods cover the measurement of average grain size and include the comparison procedure,the planimet-ric(or Jeffries)procedure,and the intercept procedures.These test methods may also be applied to nonmetallic materials with structures having appearances similar to those of the metallic structures shown in the comparison charts.These test methods apply chiefly to single phase grain structures but they can be applied to determine the average size of a particular type of grain structure in a multiphase or multiconstituent specimen.1.2These test methods are used to determine the average grain size of specimens with a unimodal distribution of grain areas,diameters,or intercept lengths.These distributions are approximately log normal.These test methods do not cover methods to characterize the nature of these distributions. Characterization of grain size in specimens with duplex grain size distributions is described in Test Methods E1181.Mea-surement of individual,very coarse grains in afine grained matrix is described in Test Methods E930.1.3These test methods deal only with determination of planar grain size,that is,characterization of the two-dimensional grain sections revealed by the sectioning plane. Determination of spatial grain size,that is,measurement of the size of the three-dimensional grains in the specimen volume,is beyond the scope of these test methods.1.4These test methods describe techniques performed manually using either a standard series of graded chart images for the comparison method or simple templates for the manual counting methods.Utilization of semi-automatic digitizing tablets or automatic image analyzers to measure grain size is described in Test Methods E1382.1.5These test methods deal only with the recommended test methods and nothing in them should be construed as defining or establishing limits of acceptability orfitness of purpose of the materials tested.1.6The measured values are stated in SI units,which are regarded as standard.Equivalent inch-pound values,when listed,are in parentheses and may be approximate.1.7This standard does not purport to address all of the safety concerns,if any,associated with its use.It is the responsibility of the user of this standard to establish appro-priate safety and health practices and determine the applica-bility of regulatory limitations prior to use.1.8The paragraphs appear in the following order:Section Number Scope1 Referenced Documents2 Terminology3 Significance and Use4 Generalities of Application5 Sampling6 Test Specimens7 Calibration8 Preparation of Photomicrographs9 Comparison Procedure101These test methods are under the jurisdiction of ASTM Committee E04onMetallography and are the direct responsibility of Subcommittee E04.08on GrainSize.Current edition approved Nov.1,2004.Published November2004.Originallyapproved st previous edition approved1996as E112–96e3.Copyright©ASTM International,100Barr Harbor Drive,PO Box C700,West Conshohocken,PA19428-2959,United States.Planimetric (Jeffries)Procedure 11General Intercept Procedures 12Heyn Linear Intercept Procedure 13Circular Intercept Procedures 14Hilliard Single-Circle Procedure 14.2Abrams Three-Circle Procedure 14.3Statistical Analysis15Specimens with Non-equiaxed Grain Shapes16Specimens Containing Two or More Phases or Constituents 17Report18Precision and Bias 19Keywords 20Annexes:Basis of ASTM Grain Size NumbersAnnex A1Equations for Conversions Among Various Grain Size Measurements AnnexA2Austenite Grain Size,Ferritic and Austenitic Steels AnnexA3Fracture Grain Size Method AnnexA4Requirements for Wrought Copper and Copper-Base Alloys AnnexA5Application to Special Situations AnnexA6Appendixes:Results of Interlaboratory Grain Size DeterminationsAppen-dix X1Referenced Adjuncts Appen-dix X22.Referenced Documents 2.1ASTM Standards:2E 3Practice for Preparation of Metallographic Specimens E 7Terminology Relating to MetallographyE 407Practice for Microetching Metals and AlloysE 562Practice for Determining V olume Fraction by Sys-tematic Manual Point CountE 691Practice for Conducting an Interlaboratory Study to Determine the Precision of a Test MethodE 883Guide for Reflected-Light PhotomicrographyE 930Test Methods for Estimating the Largest Grain Ob-served in a Metallographic Section (ALA Grain Size)E 1181Test Methods for Characterizing Duplex Grain Sizes E 1382Test Methods for Determining Average Grain Size Using Semiautomatic and Automatic Image Analysis 2.2ASTM Adjuncts:2.2.1For a complete adjunct list,see Appendix X23.Terminology3.1Definitions —For definitions of terms used in these test methods,see Terminology E 7.3.2Definitions of Terms Specific to This Standard:3.2.1ASTM grain size number —the ASTM grain size number,G ,was originally defined as:N AE 52G 21(1)where N AE is the number of grains per square inch at 100X magnification.To obtain the number per square millimetre at 1X,multiply by 15.50.3.2.2grain —that area within the confines of the original(primary)boundary observed on the two-dimensional plane-of-polish or that volume enclosed by the original (primary)boundary in the three-dimensional object.In materials contain-ing twin boundaries,the twin boundaries are ignored,that is,the structure on either side of a twin boundary belongs to the grain.3.2.3grain boundary intersection count —determination of the number of times a test line cuts across,or is tangent to,grain boundaries (triple point intersections are considered as 1-1⁄2intersections).3.2.4grain intercept count —determination of the number of times a test line cuts through individual grains on the plane of polish (tangent hits are considered as one half an interception;test lines that end within a grain are considered as one half an interception).3.2.5intercept length —the distance between two opposed,adjacent grain boundary intersection points on a test line segment that crosses the grain at any location due to random placement of the test line.3.3Symbols:Symbols:a =matrix grains in a two phase (constituent)microstructure.A =test area.A —=mean grain cross sectional area.AI ,=grain elongation ratio or anisotropy index for a longitudinally oriented plane.d —=mean planar grain diameter (Plate III).D —=mean spatial (volumetric)grain diameter.f =Jeffries multiplier for planimetric method.G =ASTM grain size number.,=mean lineal intercept length.,—a =mean lineal intercept length of the a matrix phase in a two phase (constituent)microstructure.,—,=mean lineal intercept length on a longitu-dinally oriented surface for a non-equiaxed grain structure.,—t =mean lineal intercept length on a trans-versely oriented surface for a non-equiaxed grain structure.,—p =mean lineal intercept length on a planar oriented surface for a non-equiaxed grain structure.,0=base intercept length of 32.00mm for defining the relationship between G and ,(and N L )for macroscopically or micro-scopically determined grain size by the intercept method.L =length of a test line.M =magnification used.M b =magnification used by a chart picture series.n=number of fields measured.2For referenced ASTM standards,visit the ASTM website,,or contact ASTM Customer Service at service@.For Annual Book of ASTM Standards volume information,refer to the standard’s Document Summary page on the ASTMwebsite.N a=number of a grains intercepted by the testline in a two phase(constituent)micro-structure.N A=number of grains per mm2at1X.N A a=number of a grains per mm2at1X in atwo phase(constituent)microstructure. N AE=number of grains per inch2at100X.N A,=N A on a longitudinally oriented surface fora non-equiaxed grain structure.N At=N A on a transversely oriented surface for anon-equiaxed grain structure.N Ap=N A on a planar oriented surface for anon-equiaxed grain structure.N i=number of intercepts with a test line.N Inside=number of grains completely within a testcircle.N Intercepted=number of grains intercepted by the testcircle.N L=number of intercepts per unit length oftest line.N L,=N L on a longitudinally oriented surface fora non-equiaxed grain structure.N Lt=N L on a transversely oriented surface for anon-equiaxed grain structure.N Lp=N L on a planar oriented surface for anon-equiaxed grain structure.P i=number of grain boundary intersectionswith a test line.P L=number of grain boundary intersectionsper unit length of test line.P L,=P L on a longitudinally oriented surface fora non-equiaxed grain structure.P Lt=P L on a transversely oriented surface for anon-equiaxed grain structure.P Lp=P L on a planar oriented surface for anon-equiaxed grain structure.Q=correction factor for comparison chartratings using a non-standard magnifica-tion for microscopically determined grainsizes.Q m=correction factor for comparison chartratings using a non-standard magnifica-tion for macroscopically determined grainsizes.s=standard deviation.S V=grain boundary surface area to volumeratio for a single phase structure.S V a=grain boundary surface area to volumeratio for a two phase(constituent)struc-ture.t=students’t multiplier for determination ofthe confidence interval.V V a=volume fraction of the a phase in a twophase(constituent)microstructure.95%CI=95%confidence interval.%RA=percent relative accuracy.4.Significance and Use consisting entirely,or principally,of a single phase.The test methods may also be used for any structures having appear-ances similar to those of the metallic structures shown in the comparison charts.The three basic procedures for grain size estimation are:4.1.1Comparison Procedure—The comparison procedure does not require counting of either grains,intercepts,or intersections but,as the name suggests,involves comparison of the grain structure to a series of graded images,either in the form of a wall chart,clear plastic overlays,or an eyepiece reticle.There appears to be a general bias in that comparison grain size ratings claim that the grain size is somewhat coarser (1⁄2to1G number lower)than it actually is(see X1.3.5). Repeatability and reproducibility of comparison chart ratings are generally61grain size number.4.1.2Planimetric Procedure—The planimetric method in-volves an actual count of the number of grains within a known area.The number of grains per unit area,N A,is used to determine the ASTM grain size number,G.The precision of the method is a function of the number of grains counted.A precision of60.25grain size units can be attained with a reasonable amount of effort.Results are free of bias and repeatability and reproducibility are less than60.5grain size units.An accurate count does require marking off of the grains as they are counted.4.1.3Intercept Procedure—The intercept method involves an actual count of the number of grains intercepted by a test line or the number of grain boundary intersections with a test line,per unit length of test line,used to calculate the mean lineal intercept length,,—.,—is used to determine the ASTM grain size number,G.The precision of the method is a function of the number of intercepts or intersections counted.A preci-sion of better than60.25grain size units can be attained with a reasonable amount of effort.Results are free of bias; repeatability and reproducibility are less than60.5grain size units.Because an accurate count can be made without need of marking off intercepts or intersections,the intercept method is faster than the planimetric method for the same level of precision.4.2For specimens consisting of equiaxed grains,the method of comparing the specimen with a standard chart is most convenient and is sufficiently accurate for most commer-cial purposes.For higher degrees of accuracy in determining average grain size,the intercept or planimetric procedures may be used.The intercept procedure is particularly useful for structures consisting of elongated grains.4.3In case of dispute,the intercept procedure shall be the referee procedure in all cases.4.4No attempt should be made to estimate the average grain size of heavily cold-worked material.Partially recrystallized wrought alloys and lightly to moderately cold-worked material may be considered as consisting of non-equiaxed grains,if a grain size measurement is necessary.4.5Individual grain measurements should not be made based on the standard comparison charts.These charts werethree-dimensional array of grains.Because they show a distri-bution of grain dimensions,ranging from very small to very large,depending on the relationship of the planar section and the three-dimensional array of grains,the charts are not applicable to measurement of individual grains.5.Generalities of Application5.1It is important,in using these test methods,to recognize that the estimation of average grain size is not a precise measurement.A metal structure is an aggregate of three-dimensional crystals of varying sizes and shapes.Even if all these crystals were identical in size and shape,the grain cross sections,produced by a random plane(surface of observation) through such a structure,would have a distribution of areas varying from a maximum value to zero,depending upon where the plane cuts each individual crystal.Clearly,no twofields of observation can be exactly the same.5.2The size and location of grains in a microstructure are normally completely random.No nominally random process of positioning a test pattern can improve this randomness,but random processes can yield poor representation by concentrat-ing measurements in part of a specimen.Representative implies that all parts of the specimen contribute to the result, not,as sometimes has been presumed,thatfields of average grain size are selected.Visual selection offields,or casting out of extreme measurements,may not falsify the average when done by unbiased experts,but will in all cases give a false impression of high precision.For representative sampling,the area of the specimen is mentally divided into several equal coherent sub-areas and stage positions prespecified,which are approximately at the center of each sub-area.The stage is successively set to each of these positions and the test pattern applied blindly,that is,with the light out,the shutter closed,or the eye turned away.No touch-up of the position so selected is allowable.Only measurements made onfields chosen in this way can be validated with respect to precision and bias.6.Sampling6.1Specimens should be selected to represent average conditions within a heat lot,treatment lot,or product,or to assess variations anticipated across or along a product or component,depending on the nature of the material being tested and the purpose of the study.Sampling location and frequency should be based upon agreements between the manufacturers and the users.6.2Specimens should not be taken from areas affected by shearing,burning,or other processes that will alter the grain structure.7.Test Specimens7.1In general,if the grain structure is equiaxed,any specimen orientation is acceptable.However,the presence of an equiaxed grain structure in a wrought specimen can only be determined by examination of a plane of polish parallel to the deformation axis.7.2If the grain structure on a longitudinally oriented speci-the test method.If the grain structure is not equiaxed,but elongated,then grain size measurements on specimens with different orientations will vary.In this case,the grain size should be evaluated on at least two of the three principle planes,transverse,longitudinal,and planar(or radial and transverse for round bar)and averaged as described in Section 16to obtain the mean grain size.If directed test lines are used, rather than test circles,intercept counts on non-equiaxed grains in plate or sheet type specimens can be made using only two principle test planes,rather than all three as required for the planimetric method.7.3The surface to be polished should be large enough in area to permit measurement of at leastfivefields at the desired magnification.In most cases,except for thin sheet or wire specimens,a minimum polished surface area of160mm2(0.25 in.2)is adequate.7.4The specimen shall be sectioned,mounted(if neces-sary),ground,and polished according to the recommended procedures in Practice E3.The specimen shall be etched using a reagent,such as listed in Practice E407,to delineate most,or all,of the grain boundaries(see also Annex A3).8.Calibration8.1Use a stage micrometer to determine the true linear magnification for each objective,eyepiece and bellows,or zoom setting to be used within62%.8.2Use a ruler with a millimetre scale to determine the actual length of straight test lines or the diameter of test circles used as grids.9.Preparation of Photomicrographs9.1When photomicrographs are used for estimating the average grain size,they shall be prepared in accordance with Guide E883.parison Procedure10.1The comparison procedure shall apply to completely recrystallized or cast materials with equiaxed grains.10.2When grain size estimations are made by the more TABLE1Suggested Comparison Charts for Metallic Materials N OTE1—These suggestions are based upon the customary practices in industry.For specimens prepared according to special techniques,the appropriate comparison standards should be selected on a structural-appearance basis in accordance with8.2.Material Plate Number Basic Magnification Aluminum I100X Copper and copper-base alloys(seeAnnex A4)III or IV75X,100X Iron and steel:Austenitic II or IV100X Ferritic I100X Carburized IV100X Stainless II100X Magnesium and magnesium-base alloys I or II100XNickel and nickel-base alloys II100XSuper-strength alloys I or II100XZinc and zinc-base alloys I or II100Xappearance of the standard reasonably well approaches that of the sample,errors may occur.To minimize such errors,the comparison charts are presented in four categories as follows:310.2.1Plate I —Untwinned grains (flat etch).Includes grain size numbers 00,0,1⁄2,1,11⁄2,2,21⁄2,3,31⁄2,4,41⁄2,5,51⁄2,6,61⁄2,7,71⁄2,8,81⁄2,9,91⁄2,10,at 100X.10.2.2Plate II —Twinned grains (flat etch).Includes grain size numbers,1,2,3,4,5,6,7,8,at 100X.10.2.3Plate III —Twinned grains (contrast etch).Includes nominal grain diameters of 0.200,0.150,0.120,0.090,0.070,0.060,0.050,0.045,0.035,0.025,0.020,0.015,0.010,0.005mm at 75X.10.2.4Plate IV —Austenite grains in steel (McQuaid-Ehn).Includes grain size numbers 1,2,3,4,5,6,7,8,at 100X.10.3Table 1lists a number of materials and the comparison charts that are suggested for use in estimating their average grain sizes.For example,for twinned copper and brass with a contrast etch,use Plate III.N OTE 1—Examples of grain-size standards from Plates I,II,III,and IV are shown in Fig.1,Fig.2,Fig.3,and Fig.4.10.4The estimation of microscopically-determined grain size should usually be made by direct comparison at the same magnification as the appropriate chart.Accomplish this by comparing a projected image or a photomicrograph of a representative field of the test specimen with the photomicro-graphs of the appropriate standard grain-size series,or with suitable reproductions or transparencies of them,and select thephotomicrograph which most nearly matches the image of the test specimen or interpolate between two standards.Report this estimated grain size as the ASTM grain size number,or grain diameter,of the chart picture that most closely matches the 3Plates I,II,III,and IV are available from ASTM Headquarters.Order Adjunct:ADJE11201P (Plate I),ADJE11202P (Plate II),ADJE11203P (Plate III),andFIG.1Example of Untwinned Grains (Flat Etch)from Plate I.Grain Size No.3at100XFIG.2Example of Twin Grains (Flat Etch)from Plate II.GrainSize No.3at100XFIG.3Example of Twin Grains (Contrast Etch)from Plate III.Grain Size 0.090mm at75X10.5Good judgment on the part of the observer is necessary to select the magnification to be used,the proper size of area (number of grains),and the number and location in the specimen of representative sections and fields for estimating the characteristic or average grain size.It is not sufficient to visually select what appear to be areas of average grain size.Recommendations for choosing appropriate areas for all pro-cedures have been noted in 5.2.10.6Grain size estimations shall be made on three or more representative areas of each specimen section.10.7When the grains are of a size outside the range covered by the standard photographs,or when magnifications of 75X or 2and Table 2.It may be noted that alternative magnifications are usually simple multiples of the basic magnifications.N OTE 2—If the grain size is reported in ASTM numbers,it is conve-nient to use the relationship:Q 52log 2~M /M b !(2)56.64log 10~M /M b !where Q is a correction factor that is added to the apparent micro-grain size of the specimen,as viewed at the magnification,M ,instead of at the basic magnification,M b (75X or 100X),to yield the true ASTM grain-size number.Thus,for a magnification of 25X,the true ASTM grain-size number is four numbers lower than that of the corresponding photomi-crograph at 100X (Q =−4).Likewise,for 400X,the true ASTM grain-sizeFIG.4Example of Austenite Grains in Steel from Plate IV.GrainSize No.3at 100XTABLE 2Microscopically Determined Grain Size Relationships Using Plate III at Various MagnificationsN OTE 1—First line—mean grain diameter,d,in mm;in parentheses—equivalent ASTM grain size number,G.N OTE 2—Magnification for Plate III is 75X (row 3data).MagnificationChart Picture Number (Plate III)123456789101112131425X 0.015(9.2)0.030(7.2)0.045(6.0)0.060(5.2)0.075(4.5)0.105(3.6)0.135(2.8)0.150(2.5)0.180(2.0)0.210(1.6)0.270(0.8)0.360(0)0.451(0/00)0.600(00+)50X 0.0075(11.2)0.015(9.2)0.0225(8.0)0.030(7.2)0.0375(6.5)0.053(5.6)0.0675(4.8)0.075(4.5)0.090(4.0)0.105(3.6)0.135(2.8)0.180(2.0)0.225(1.4)0.300(0.5)75X 0.005(12.3)0.010(10.3)0.015(9.2)0.020(8.3)0.025(7.7)0.035(6.7)0.045(6.0)0.050(5.7)0.060(5.2)0.070(4.7)0.090(4.0)0.120(3.2)0.150(2.5)0.200(1.7)100X 0.00375(13.2)0.0075(11.2)0.0112(10.0)0.015(9.2)0.019(8.5)0.026(7.6)0.034(6.8)0.0375(6.5)0.045(6.0)0.053(5.6)0.067(4.8)0.090(4.0)0.113(3.4)0.150(2.5)200X 0.0019(15.2)0.00375(13.2)0.0056(12.0)0.0075(11.2)0.009(10.5)0.013(9.6)0.017(8.8)0.019(8.5)0.0225(8.0)0.026(7.6)0.034(6.8)0.045(6.0)0.056(5.4)0.075(4.5)400X —0.0025(14.3)0.0037(13.2)0.005(12.3)0.006(11.7)0.009(10.7)0.011(10.0)0.0125(9.7)0.015(9.2)0.0175(8.7)0.0225(8.0)0.030(7.2)0.0375(6.5)0.050(5.7)500X——0.003(13.8)0.004(13.0)0.005(12.3)0.007(11.4)0.009(10.6)0.010(10.3)0.012(9.8)0.014(9.4)0.018(8.6)0.024(7.8)0.030(7.2)0.040(6.3)number is four numbers higher than that of the corresponding photomi-crograph at 75X.10.8The small number of grains per field at the coarse end of the chart series,that is,size 00,and the very small size of the grains at the fine end make accurate comparison ratings difficult.When the specimen grain size falls at either end of the chart range,a more meaningful comparison can be made by changing the magnification so that the grain size lies closer to the center of the range.10.9The use of transparencies 4or prints of the standards,with the standard and the unknown placed adjacent to each other,is to be preferred to the use of wall chart comparison with the projected image on the microscope screen.10.10No particular significance should be attached to the fact that different observers often obtain slightly different results,provided the different results fall within the confidence limits reasonably expected with the procedure used.10.11There is a possibility when an operator makes re-peated checks on the same specimen using the comparison method that they will be prejudiced by their first estimate.This disadvantage can be overcome,when necessary,by changes in magnification,through bellows extension,or objective or eyepiece replacement between estimates (1).510.12Make the estimation of macroscopically-determined grain sizes (extremely coarse)by direct comparison,at a magnification of 1X,of the properly prepared specimen,or of a photograph of a representative field of the specimen,with photographs of the standard grain series shown in Plate I (for untwinned material)and Plates II and III (for twinned mate-rial).Since the photographs of the standard grain size series were made at 75and 100diameters magnification,grain sizes estimated in this way do not fall in the standard ASTM grain-size series and hence,preferably,should be expressed4Transparencies of the various grain sizes in Plate I are available from ASTM Headquarters.Order Adjunct:ADJE112TS for the set.Transparencies of individual grain size groupings are available on request.Order Adjunct:ADJE11205T (Grain Size 00),ADJE11206T (Grain Size 0),ADJE11207T (Grain Size 0.5),ADJE11208T (Grain Size 1.0),ADJE11209T (Grain Size 1.5),ADJE11210T (Grain Size 2.0),ADJE11211T (Grain Size 2.5),ADJE11212T (Grain Sizes 3.0, 3.5,and 4.0),ADJE11213T (Grain Sizes 4.5,5.0,and 5.5),ADJE11214T (Grain Sizes 6.0,6.5,and 7.0),ADJE11215T (Grain Sizes 7.5,8.0,and 8.5),and ADJE11216T (Grain Sizes 9.0,9.5,and 10.0).Charts illustrating grain size numbers 00to 10are on 8⁄TABLE 3Macroscopic Grain Size Relationships Computed for Uniform,Randomly Oriented,Equiaxed GrainsN OTE 1—Macroscopically determined grain size numbers M-12.3,M-13.3,M-13.8and M-14.3correspond,respectively,to microscopically determined grain size numbers (G )00,0,0.5and 1.0.Macro Grain Size No.N ¯A Grains/Unit Area A ¯Average Grain Area d —Average Diameter ,—Mean Intercept N ¯L N ¯No./mm 2No./in.2mm 2in.2mm in.mm in.mm −1100mm M-00.00080.501290.3 2.0035.9 1.4132.00 1.20.031 3.13M-0.50.00110.71912.4 1.4130.2 1.1926.91 1.00.037 3.72M-1.00.0016 1.00645.2 1.0025.4 1.0022.630.890.044 4.42M-1.50.0022 1.41456.20.70721.40.84119.030.740.053 5.26M-2.00.0031 2.00322.60.50018.00.70716.000.630.063 6.25M-2.50.0044 2.83228.10.35415.10.59513.450.530.0747.43M-3.00.0062 4.00161.30.25012.70.50011.310.440.0888.84M-3.50.0088 5.66114.00.17710.70.4209.510.370.10510.51M-4.00.01248.0080.640.1258.980.3548.000.310.12512.50M-4.50.017511.3157.020.08847.550.297 6.730.260.14914.87M-5.00.024816.0040.320.0625 6.350.250 5.660.220.17717.68M-5.50.035122.6328.510.0442 5.340.210 4.760.180.21021.02M-6.00.049632.0020.160.0312 4.490.177 4.000.150.25025.00M-6.50.070145.2614.260.0221 3.780.149 3.360.130.29729.73M-7.00.09964.0010.080.0156 3.170.125 2.830.110.35435.36M-7.50.14090.517.130.0110 2.670.105 2.380.0930.42042.05310−3310−3310−3M-8.00.198128.0 5.047.812 2.2588.4 2.0078.70.50050.00M-8.50.281181.0 3.56 5.524 1.8974.3 1.6866.20.59559.46M-9.00.397256.0 2.52 3.906 1.5962.5 1.4155.70.70770.71M-9.50.561362.1 1.78 2.762 1.3352.6 1.1946.80.84184.09M-10.00.794512.0 1.26 1.953 1.1244.2 1.0039.4 1.00100.0M-10.5 1.122724.10.891 1.3810.99437.20.84133.1 1.19118.9M-11.0 1.5871024.10.6300.9770.79431.20.70727.8 1.41141.4M-11.5 2.2451448.20.04450.6900.66726.30.59523.4 1.68168.2M-12.0 3.1752048.10.3150.4880.56122.10.50019.7 2.00200.0M-12.3 3.9082521.60.2560.3970.50619.90.45117.7 2.22221.9M-12.5 4.4902896.50.2230.3450.47218.60.42016.6 2.38237.8M-13.0 6.3494096.30.1570.2440.39715.60.35413.9 2.83282.8M-13.37.8175043.10.1280.1980.35814.10.31912.5 3.14313.8M-13.58.9795793.00.1110.1730.33413.10.29711.7 3.36336.4M-13.811.0557132.10.0910.1400.30111.80.26810.5 3.73373.2M-14.012.6998192.60.0790.1220.28111.00.2509.84 4.00400.0M-14.315.63410086.30.0640.0990.2539.960.2258.874.44443.8。

Mathematics for Economists by Simon Blume 习题解答 answers3

Mathematics for Economists by Simon  Blume 习题解答 answers3

ANSWERS PAMPHLET
103
0 2 x Ϫ1 18.3): det H ϭ det 2x 2 ϩ 2␭ 0 ϭ Ϫ(2 ϩ 2␭ ϩ 8 x 2 ) Ͻ 0, Ϫ1 0 2 the SOC for a constrained min. 0 0 3 1 1 0 0 1 1 1 18.5): 3 1 2 0 0 1 1 0 2 0 1 1 0 0 2 has positive determinant, the SOC for a constrained min when there are 3 variables and 2 constraints. 19.15 If
␮j‫( ء‬a)h j (x‫( ء‬a); a)
ϭ L(x‫( ء‬a), ␮ ‫( ء‬a); a) for all a. d d f (x‫( ء‬a)) ϭ L(x‫ ء‬, ␮ ‫( ء‬a); a) da da dx ‫ء‬ ‫ץ‬L ‫ء‬ (x (a), ␮ ‫( ء‬a), a) i (a) ϭ ‫ץ‬xi da i ϩ
The Jacobian for these expressions with respect to (x1 , x2 , ␭ ) is 2x2 Ϫ 4␭ 2 x1 Ϫ4 x1 2 x1 Ϫ2␭ Ϫ2 x2 Ϫ4 x1 Ϫ2 x2 . 0
At x1 ϭ x2 ϭ 1 and ␭ ϭ 0.5, its determinant is 48. The implicit function theorem concludes that we can solve L x1 ϭ 0, L x2 ϭ 0, L␭ ϭ 0 for (x1 , x2 , ␭ ) as C 1 functions of a near a ϭ 3. 19.19 If Dh(x)‫ ء‬does not have maximal rank, we can use elementary row operations to transform one of the last k rows of D2 Lx,␮ in (31) to a zero row. This would imply that D2 Lx,␮ itself does not have maximal rank. 19.20 The new Lagrangian is

欧洲道路标准

欧洲道路标准

欧洲道路标准集团档案编码:[YTTR-YTPT28-YTNTL98-UYTYNN08]E N13201-3Edition: 2007-06-01路灯第三部分:性能计算Road lightingPart 3: Calculation of performanceNational ForewordThe present ?NORM EN has been reissued without prior public enquiry and represents a consolidated national new edition of EN 13201-3:2003-11, including Corrigendum EN 13201-3:2003/AC:2007-02.The preceding European corrigendum EN 13201-3:2003/AC:2005-06 has been considered and incorporated in this version of ?NORM EN 13201-3.Following clauses were changed:Ad Figure 11 was replacedText after figure title was addedAd Equation (35) was includedEnglish versionRoad lighting - Part 3: Calculation of performanceThis European Standard was approved by CEN on 1 September 2003.This corrigendum becomes effective on 22 June 2005.This corrigendum becomes effectice on 28 February 2007.CEN members are bound to comply with the CEN/CENELEC Internal Regulations which stipulate theconditions for giving this European Standard the status of a national standard without any alteration. Up-to-date lists and bibliographical references concerning such national standards may be obtained on application to the Management Centre or to any CEN member.This European Standard exists in three official versions (English, French, German). A version in any other language made by translation under the responsibility of a CEN member into its own language and notified to the Management Centre has the same status as the official versions.CEN members are the national standards bodies of Austria, Belgium, Bulgaria, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Norway, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden, Switzerland and United Kingdom.Contents pageForeword ......................................................................... .. (4)Introduction ..................................................................... . (5)1 Scope............................................................................. . (5)2 Normativereferences............ ........................................................... . (5)3 Terms, definitions, symbols and abbreviations (5)Terms and definitions....................................................................... . (5)List of symbols and abbreviations..................................................................... . (8)4 Mathematicalconventions ............ ......................................................... . (10)5 Photometric data.............................................................................. . (10)General .......................................................................... .. (10)The _-table ............................................................................ (10)Interpolation in the _-table ............................................................................ .. (12)General .......................................................................... .. (12)Linear interpolation..................................................................... (12)Quadratic interpolation..................................................................... . (13)Quadratic interpolation in the region of C = 0_, or _ = 0_ or180_ (15)The r-table ............................................................................ (15)Interpolation in the r-table............................................................................. . (18)6 Calculation of_(C,_) ........................................................................... (19)General .......................................................................... .. (19)Mathematical conventions for distances measured on the road (19)Mathematical conventions for rotations (20)Calculation of C and_ ................................................................................ .. (21)7 Calculation of photometric quantities........................................................................ (22)Luminance ........................................................................ . (22)Luminance at apoint ............................................................................ .. (22)Total luminance at a point............................................................................. .. (23)Field of calculation for luminance......................................................................... .. (23)Position of calculationpoints ........................................................................... (24)llluminance....................................................................... .. (29)General .......................................................................... .. (29)Horizontal illuminance at a point............................................................................. .. 29Hemispherical illuminance at a point (29)Semicylindrical illuminance at apoint (30)Vertical iluminance at apoint ............................................................................ (31)Total illuminance at a point............................................................................. (32)Field of calculation for illuminance....................................................................... .. (33)Position of calculationpoints ........................................................................... .. (33)Luminaires included incalculation ...................................................................... (34)llluminance on areas of irregularshape (35)8 Calculation of qualitycharacteristics .................................................................. .. (35)General .......................................................................... .. (35)Averageluminance ........................................................................ .. (35)Overall uniformity........................................................................ .. (35)Longitudinaluniformity ....................................................................... . (35)Threshold increment......................................................................... .. (35)Surroundratio ............................................................................ .. (36)Measures ofilluminance ...................................................................... (39)General .......................................................................... . (39)Averageilluminance ...................................................................... (39)Minimum illuminance....................................................................... .. (40)Uniformity ofilluminance ...................................................................... . (40)9 Ancillary data.............................................................................. .. (40)Bibliography ..................................................................... (41)ForewordThis document (EN 13201-3:2003) has been prepared by Technical Committee CEN/TC 169 “Light and lighting”, the secretariat of which is held by DIN.This European Standard shall be given the status of a national standard, either by publication of an identical text or by endorsement, at the latest by May 2004, and conflicting national standards shall be withdrawn at the latest by May 2004.This European Standard was worked out by the Joint Working Group of CEN/TC 169 "Light and lighting" and CEN/TC 226 "Road Equipment", the secretariat of which is held by AFNOR.This document includes a Bibliography.This standard, EN 13201 Road lighting, consists of three parts. This document is:Part 3: Calculation of performanceThe other parts of EN 13201 are:Part 2: Performance requirementsPart 4: Methods of measuring lighting performanceAccording to the CEN/CENELEC Internal Regulations, the national standards organizations of thefollowing countries are bound to implement this European Standard: Austria, Belgium, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Luxembourg, Malta, Netherlands, Norway, Portugal, Slovakia, Spain, Sweden, Switzerland and the United Kingdom.IntroductionThe calculation methods described in this Part of this European Standard enable road lighting quality characteristics to be calculated by agreed procedures so that results obtained from different sources will have a uniform basis.引言本部分欧洲标准中描述的计算方法能使路灯指标通过认可的程序计算出,以使不同光源的结果具有统一的计算依据。

FRM-STOCHASTIC PROCESSES FOR FINANCE RISK Management TOOLS

FRM-STOCHASTIC PROCESSES FOR FINANCE RISK Management TOOLS

STOCHASTIC PROCESSES FOR FINANCE RISK MANAGEMENT TOOLSNotes for the Course byF.Boshuizen,A.W.van der Vaart,H.van Zanten and K.Banachewicz(PARTLY)CORRECTED DRAFT,September2005iiCONTENTS1.Pricing derivatives (1)1.1.Hedging a Forward (1)2.Binomial Tree Model (4)2.1.One Period Model (4)2.2.Two Period Model (6)2.3.N Period Model (7)3.Discrete Time Stochastic Processes (8)3.1.Stochastic Processes (8)3.2.Conditional Expectation (9)3.3.Filtration (10)3.4.Martingales (11)3.5.Change of Measure (11)3.6.Martingale Representation (12)4.Binomial Tree Model Revisited (13)4.1.Towards Continuous Time (16)5.Continuous Time Stochastic Processes (18)5.1.Stochastic Processes (18)5.2.Brownian Motion (18)5.3.Filtrations (20)5.4.Martingales (20)5.5.Generalized Brownian Motion (20)5.6.Variation (21)5.7.Stochastic Integrals (23)5.8.Geometric Brownian Motion (24)5.9.Stochastic Differential Equations (24)5.10.Markov Processes (25)5.11.Quadratic variation-revisited (25)5.12.Itˆo Formula (27)5.13.Girsanov’s Theorem (29)5.14.Brownian Representation (30)6.Black-Scholes Model (32)6.1.Portfolios (33)6.2.The Fair Price of a Derivative (33)6.3.European Options (35)6.4.The Black-Scholes PDE and Hedging (36)6.5.The Greeks (38)6.6.General Claims (39)6.7.Exchange Rate Derivatives (39)7.Extended Black-Scholes Models (41)7.1.Market Price of Risk (42)7.2.Fair Prices (42)7.3.Arbitrage (44)7.4.PDEs (45)iii 8.Interest Rate Models (48)8.1.The Term Structure of Interest Rates (48)8.2.Short Rate Models (50)8.3.The Hull-White Model (54)8.4.Pricing Interest Rate Derivatives (57)8.5.Examples of Interest Rate Derivatives (59)9.Risk Measurement (63)9.1.Value-At-Risk (63)9.2.Normal Returns (65)9.3.Equity Portfolios (66)9.4.Portfolios with Stock Options (68)9.5.Bond Portfolios (70)9.6.Portfolios of Bonds and Swaptions (71)9.7.Diversified Portfolios (73)ivLITERATURE[1]Baxter,M.and Rennie,A.,(1996).Financial calculus.CambridgeUniversity Press,Cambridge.[2]Chung,K.L.and Williams,R.J.,(1990).Introduction to stochasticintegration,second edition.Birkh¨a user,London.[3]Etheridge,A.,(2002).A Course in Financial Calculus.CambridgeUniversity Press.[4]Campbell,J.Y.,Lo,A.W.and MacKinlay,A.C.,(1997).The Econo-metrics of Financial Markets.Princeton University Press.[5]Hunt,P.J.and Kennedy,J.E.,(1998).Financial Engineering.Wiley.[6]Jorion,P.,(2001).Value at Risk the New Benchmark for ManagingFinancial Risk.McGraw-Hill,New York.[7]Musiela,M.and Rutkowski,M.,(1997).Martingale Methods in Fi-nancial Modelling.Springer-Verlag,Berlin.[8]Smithson,C.W.,Smith,C.W.and Wilford,D.S.,(1995).ManagingFinancial Risk.Irwin,Burr Ridge,Illinois.Baxter and Rennie is a book written for an audience of people in practice, but using the correct mathematical concepts and results.The result is a text with much intuition,and with mathematical theorems that are stated with some precision,but never proved or rigorously interpreted.Etheridge is a more mathematical version of Baxter and Rennie.Together these books are close to the content of the course.We recommend that you read these books,next to the following notes,which are really very brief.The other books on the list are for further reading.Musiela and Rutkowski is a step up in mathematical level.Hunt and Kennedy and Chung and Williams are mathematically completely rigorous. Chung and Williams has very little onfinance.This is just a tiny selection of the books on mathematicalfinance that have appeared in the past ten years.Campbell,Lo and MacKinLay gives a much wider view onfinance, including some historical analysis and economic theories of price form-ing through utility,next to subjects close this course,all from a low-level mathematics point of view.The authors of“Managing Financial Risk”are bankers,not mathematicians.In their preface they write:This stuffis not as hard as some people make it sound.Thefinancial markets have some complicated features,but good com-mon sense goes a lot further than mathematicalflash and dash.Keep that mind when following this course.1Pricing derivativesFinancial instruments can be divided in two basic classes:underlying and derivative ones.The former can be stocks,bonds or various trade goods, while the latter arefinancial contracts that promise some payment to the owner,depending on the behavior of the“underlying”(e.g.a price at a given time T or an average price over a certain time period).Derivatives are extremely useful for risk management(apart from obvious investment properties)—we can reduce ourfinancial vulnerability byfixing a price for a future transaction now.In this chapter we introduce some basic concepts through examples; formal definitions and theory follows in later chapters.1.1Hedging a ForwardA forward is a contract that pays the owner an amount S T−K at afixed time T in the future,the expiry time,where S t is the price of an asset at time t and K is afixed number,the strike price.Both T and K are written in the contract,but S T will be known only at the expiry time.For which number K is the value of this contract equal to0at time0?Suppose that we have,besides buying the contract,two other options to invest our money:(i)We can put our money in a savings account against afixed,predeter-mined interest rate r.One unit of money placed in the account grows to e rt units during a time interval[0,t]and is freely available.A neg-ative balance in our account is permitted,thus allowing us to borrow money at the same interest rate r.If we borrow one unit at time0, than we owe e rt units at time t,which is equivalent to having a capital of−e rt units.21:Pricing derivatives(ii)We can invest in the asset.The asset price S t at time t is a stochasticvariable,dependent on t .It may be assumed that we know the proba-bility distributions of these variables.For instance,a popular model is that the variable S T /S 0is log normally distributed.We discuss two answers to the pricing question.A naive (wrong)answer is to argue as follows.A payment of S T −K at time T is worth e −rT (S T −K )at time 0.This is an unknown stochastic quantity from the perspective of time 0.The reasonable price is the expec-tation E e −rT (S T −K ) of this variable.The strike price that gives the value 0is therefore the solution of the equation0=E e −rT (S T −K )=e −rT (E S T −K ).In other words,K =E S T .For instance,using a log normal distribution with parameters µT and σ2T for S T /S 0,i.e.log(S T /S 0)is normally distributed with these parameters,we find thatK =E S T =S 0e µT +12σ2T.Somewhat surprisingly,this does not depend on the interest rate,meaning that somehow no trade-offhas been made between investing in the asset or putting money in the savings account.An accepted solution to the pricing question compares buying the con-tract to a “hedging strategy”,as follows:(i)at time 0borrow an amount S 0at interest rate r ,and buy the assetat price S 0,(ii)sit still until expiry time T .At time T we own the asset,worth S T ,and we owe S 0e rT to the bank.If K =S 0e rT ,then together this is exactly the amount S T −K paid by the forward contract at time T .Thus in that case we ought to be indifferent towards buying the forward or carrying out the strategy (i)-(ii).The strat-egy costs us nothing at time 0and hence K =S 0e rT is the correct strike price for a forward with value 0at time 0;this is also a maximal price,that the buyer of the contract can accept.Since the seller of the contract can perform a symmetric reasoning with S 0e rT as a minimal accepted price,K =S 0e rT is the unique fair price acceptable to both parties.The correct solution K =S 0e rT does not depend on the distribu-tion of the asset price S T ,and in fact depends on the asset price process only through its (observable)value at time 0.This is remarkable if you are used to evaluate future gains through expectations on the random vari-ables involved.Is there no role for probability theory in evaluating financial contracts?There is.First we note that the expected gain of owning a contract is equal to E(S T −K ),which does depend on the distribution of the asset price.It evaluates toS 0(e µT +12σ2T−e rT )1.1:Hedging a Forward3 if the asset price is log normally distributed and K=S0e rT.Second,it turns out that the correct solution can in fact be found from computing an expectation,but the expectation should be computed under a special probability measure,called a“martingale measure”.To evaluate the price of a forward,this route would be overdone,as the preceding hedging strategy is explicit and simple.However,the prices of other contracts may not be so easy to evaluate,and in general probabilities and expectations turn out to be very useful.For instance,consider a European put option, which is a contract that pays an amount(S T−K)+at time T,i.e.S T−K if S T>K and nothing otherwise.There is no simple universal hedging strategy to price this contract,but it turns out that given reasonable prob-abilistic models for the asset price process S t,there are more complicated trading strategies that allows to parallel the preceding reasoning.These strategies require continuous trading during the term[0,T]of the contract, and a big mathematical apparatus for its evaluation.2Binomial Tree ModelAfinancial derivative is a contract that is based on the price of an underly-ing asset,such as a stock price or bond price.An“option”,of which there are many different types,is an important example.A main objective of “financial engineering”is tofind a“fair price”of such a derivative,where by.fair we mean a price acceptable both for the buyer and the seller.Fol-lowing the work by Black and Scholes in the1970s the prices of derivatives are found through the principle of“no arbitrage”introduced at the end of the previous chapter.In this chapter we discuss,as an introduction,the pricing of a European call option using a discrete time framework.2.1One Period ModelSuppose that at time0we can invest in an asset with price s0,or put money in a savings account with afixed interest rate.We model the asset price at time1as a random variable S1that can take only two valuesP(S1=us0)=p,P(S1=ds0)=1−p.Here u(for“up”)and d(for“down”)are two known constants,with u>d, and p is a number in[0,1]that may be unknown.We assume that an amount of one unit of money put in the savings account at time0grows to a guaranteed amount e r units at time1.We want tofind the fair price at time0of a contract that pays the amount C at time1,where C=C(S1)may depend on the(unknown)value of the asset at the payment date.2.1:One Period Model52.1Example.A European call option corresponds to C=(S1−K)+, for a given strike price K.The payment on a forward contract is equal to C=S1−K.Suppose that at time0we buyφ0assets and put an amount ofψ0 money units in the savings account.Then we have a“portfolio”(φ0,ψ0) whose worth at time0is given by(2.2)V0=φ0S0+ψ0·1.If we do not trade between the times0and1,then the value of the portfolio changes to its value V1at time1,given byV1=φ0S1+ψ0e r.The asset can only take the values us0and ds0at time1.In thefirst case the contract is worth C(us0)at time1,whereas in the second case it is worth C(ds0)at time1.The value of the portfolio is equal toφ0us0+ψ0e r orφ0ds0+ψ0e r in the two cases.Suppose that wefix the portfolio(φ0,ψ0) so that its value at time1agrees exactly with the contract,for each of the two possibilities,i.e.(2.3)φ0us0+ψ0e r=C(us0),φ0ds0+ψ0e r=C(ds0).This portfolio will cost us V0at time0,and is guaranteed to have the same value at time1as the contract with claim C(S1),whether the asset moves up or down.We should therefore have no preference for the portfolio or the contract,and hence a fair price for the contract at time0is the price of the portfolio,i.e.V0corresponding to the portfolio(φ0,ψ0)satisfying the equations(2.3).The equations(2.3)form a system of two linear equations in the un-knownsφ0andψ0and can be solved to giveφ0=C(us0)−C(ds0)us0−ds0,ψ0=e−r uC(ds)−dC(us0)u−d.Inserting this in the equation(2.2),we see that this portfolio can be acquired at time zero for the amount(2.4)V0=e−rqC(us0)+(1−q)C(ds0),q=e r−du−d.This is the fair price of the contract at time0.For d≤e r≤u the number q is contained in the interval[0,1]and can be considered an alternative probability for the upward move of the asset process.In general,this probability is different from the probability62:Binomial Tree Modelp,which turns out to be unimportant for the fair price.It can be seen that q is the unique probability such that(2.5)E q(e−r S1)=s0.Furthermore,the price of the contract can be written asV0=E qe−r C(S1).We can write equation(2.5)also in the form E q(e−r S1|S0)=S0,which expresses that the expected value of the discounted asset price e−r S1given S0is equal to S0=e−0S0,or that the process S0,e−r S1is a“martingale”.2.2Two Period ModelSuppose that at time0we have the same possibilities for investing as in the preceding section,but we now consider a full trading horizon of three times:0,1,2.We wish to evaluate a claim on the asset process payable at time2.Let the price of the asset at the three times be modelled by S0,S1,S2, where we assume that S0=s0isfixed,S1is equal to either dS0or uS0, and S2is equal to either dS1or uS1.Thus the asset prices follow a path in a binary tree.We assume that at each node of the tree the decision to move up or down is made with probabilities p and1−p,independently for the different nodes.Besides investing in the asset we may put money in a savings account (also a negative amount,indicating that we borrow money)at afixed in-terest rate r.One unit in the savings account grows to e r units at time1, and to e2r units at time2.The contract pays the amount C=C(S2)at time2,and we wish to find its fair price at time0.We can evaluate the claim recursively,backwards in time.At time2the claim is worth C.At time1there are two possibilities:either the asset price is dS0or it is uS0.If we put ourselves at the perspective of time1,then we know which of the two possibilities is realized.If s1is the realized value of S1,then we can calculate the value of the claim using the one-period modelas(cf.(2.4))e−rqC(us1)+(1−q)C(ds1).For the two possibilities for s1this givese−rqC(uds0)+(1−q)C(d2s0),if S1=ds0,e−rqC(u2s0)+(1−q)C(dus0),if S1=us0.2.3:N Period Model7 This is the value of the contract at time1,as a function of the asset price S1at time one.We can think of this value as the pay-offon our contract at time1,and next apply the one-period model a second time to see that the value of the contract at time0is given by(cf.(2.4))e−rqe−rqC(u2s0)+(1−q)C(dus0)+(1−q)e−rqC(uds0)+(1−q)C(d2s0).This equation can be rearranged ase−2rq2C(u2s0)+2q(1−q)C(uds0)+(1−q)2C(d2s0)=E qe−2r C(S2).Hence once again the price is the expectation of the discounted claim, presently e−2r C(S2),under the probability measure on the tree given by the branching probability q.2.3N Period ModelWe can price a claim in a binomial tree model with N periods by extending the backwards induction argument.The fair price of a claim C(S N)is givenbyE qe−Nr C(S N).3Discrete TimeStochastic Processes3.1Stochastic ProcessesA stochastic process in discrete time is a(finite or infinite)sequence X= (X0,X1,...)of random variables or vectors,defined on a given probability space.Mathematically,random variables are maps X n:Ω→R that map outcomesω∈Ωinto numbers X n(ω).The stochastic process X0,X1,..., maps every outcomeωinto a sequence of numbers X0(ω),X1(ω),...,called a sample path.The best way to think of a stochastic process is to visualize the sample paths as“random functions”.We generate an outcomeωaccording to some probability measure on the set of all outcomes and next have a function n→X n(ω)on the domain N∪{0}.This domain is referred as the set of “discrete times”.3.1Example(Binomial tree model).The binomial tree model for the stock price is a stochastic process S0,S1,...,S N,where each possible sam-ple path is given by a path in the binomial tree,and the probability of a sample path is the product of the probabilities on the branches along the path.As indicated before,the best way to think about this stochastic process is as a random function,generated according to the probabilities attached to the different paths on the tree.The preceding description gives an in-tuitively clear description of the binomial tree process,but for later use it is instructive to define the stochastic process also formally as a map on a given outcome space.One possibility is to takeΩto be equal to the set of N-tuplesω=(ω1,...,ωN),where eachωi∈{0,1}.The appropriate3.2:Conditional Expectation9 probability measure isP({(ω1,...,ωN)})=p#1≤i≤N:ωi=1(1−p)#1≤i≤N:ωi=0,and the stochastic process can be formally defined by setting S0=s0and, for n=1,2,...,N,S n(ω1,...,ωN)=S0u#1≤i≤n:ωi=1d#1≤i≤n:ωi=0.Thusωi=1indicates that the sample path goes up in the tree at time i, whereasωi=0indicates a down move.The value S n at time n is determined by the total number of moves up and down in the tree up till that time.3.2Conditional ExpectationFor a discrete random variable X and a discrete random vector Y,the conditional expectation of X given the event Y=y is given byE(X|Y=y)=x P(X=x|Y=y).xIf we write this function of y as f(y)=E(X|Y=y),then we write E(X|Y) for f(Y).This is a random variable,called the conditional expectation of X given Y.Some important rules are given in the following lemma.3.2Lemma.(i)EE(X|Y)=E X.(ii)E(E(X|Y,Z)|Z)=E(X|Z)(tower property).(iii)E(X|Y)=X if X=f(Y)for some function f.The three rules can be proved from the definition,but are intuitively clear.Thefirst rule says that the expectation of a variable X can be com-puted in two steps,first using the information on another variable Y,and next taking the expectation of the result.Assertion(ii)gives exactly the same property,with the difference that every of the expectations are com-puted conditionally on a variable Z.Rule(iii)says that we can predict a variable X exactly if X is a function of a known variable Y,which is obvious.We shall use the notation E(X|Y)also if X or Y are continuous ran-dom variables or vectors.Then the preceding definition does not make sense, because the probabilities P(X=x|Y=y)are not defined if P(Y=y)=0,103:Discrete Time Stochastic Processeswhich is the case for continuous random variable X.However,the condi-tional expectation E(X|Y)can still be defined as a function of Y,namely as the function such that,for every function g,EE(X|Y)g(Y)=EXg(Y).The validity of this equality in the case of discrete random variables can be checked in the same manner as the validity of the three rules in the lemma.For general random variables X and Y we take this as a definition of conditional expectation,where it is also understood that E(X|Y)must be a function of Y.The three rules of the lemma continue to hold for this extended definition of conditional expectation.3.3FiltrationAσ-field is a collection of events.Afiltration in discrete time is an increasing sequence F0⊂F1⊂···ofσ-fields,one per time instant.Theσ-field F n may be thought of as the events of which the occurrence is determined at or before time n,the“known events”at time n.Filtrations are important to us,because they allow to model theflow of information.Of course,the information increases as time goes by.Thefiltrations that are of interest to us are generated by stochastic processes.The naturalfiltration of a stochastic process X0,X1,...is defined byF n={(X0,X1,...,X n)∈B:B⊂R n+1}.Thus F n contains all events that depend on thefirst(n+1)elements of the stochastic process.It gives the“history”of the process up till time n.A convenient notation to describe aσ-field corresponding to observing a random vector X isσ(X).Thusσ(X),called theσ-field generated by X,consists of all events that can be expressed in X:events of the type {X∈B}.In this notation,the naturalfiltration of a stochastic process X0,X1,...can be written as F n=σ(X0,...,X n).We say that a process X0,X1,...is adapted to thefiltration(F n)n≥0if σ(X n)⊂F n for every n.Thus the events connected to an adapted process up to time n are known at time n.The naturalfiltration corresponding to a process is the smallestfiltration to which it is adapted.If the pro-cess Y0,Y1,...is adapted to the naturalfiltration of a stochastic process X0,X1,...,then for each n the variable Y n is a functionφn(X0,X1,...,X n) of the sample path of the process X up till time n.We say that a process X0,X1,...is predictable relative to thefiltration (F n)n≥0ifσ(X n)⊂F n−1for each n.Thus the events connected to a predictable process are known one time instant before they happen.3.4:Martingales 11If F is the σ-field generated by Y ,then we also write E(X |F )for the random variable E(X |Y ).Thus E(X |F )is the expected value of X given the information F .3.3Lemma.(i)EE(X |Y )=E X .(ii)for two σ-fields F ⊆G there holds E(E(X |G )|F )=E(X |F )(towerproperty).(iii)E(X |Y )=X if X ∈σ(Y ).3.4MartingalesA stochastic process X 0,X 1,...is a martingale relative to a given filtration (F n )n ≥0if it is adapted to this filtration and E(X n |F m )=X m for every m <n .The martingale property is equivalent to E(X n −X m |F m )=0for every m <n ,expressing that the increment X n −X m given the “past”F n has expected value 0.A martingale is a stochastic process that,on the average,given the past,does not grow or decrease.3.4Example (Doob martingale).If Y is a random variable with E |Y |<∞and F n an arbitrary filtration,then X n =E(Y |F n )defines a martingale.This can be proved from the tower property of conditional expectations:E E(Y |F n )|F m )=E(Y |F m )for any m <n .The martingale property can also be equivalently described through one step ahead expectations.A process is a martingale if E(X n +1|F n )=X n for every n .e the tower property to prove this.3.5Change of MeasureIf there are two possible probability measures P and Q on the set of out-comes,and X is a martingale relative to Q ,then typically it is not a mar-tingale relative to P .This is because the martingale property involves the expected values,and hence the probabilities of the various outcomes.An important tool in finance is to change a given measure P into a measure Q making a discounted asset process into a martingale.123:Discrete Time Stochastic Processes3.6Example.In the binomial tree model with F n the naturalfiltration of S0,S1,...and P(S n+1=uS n|F n)=1−P(S n+1=dS n|F n)=p,we havethatE(S n+1|F n)=uS n p+dS n(1−p)=S nup+d(1−p).This is equal to S n if u,d and p satisfy the equation up+d(1−p)=1.For instance,if u=2and d=1/2,then this is true if and only if p=1/3.3.7Example(Discounted stock).In the binomial tree model as in the preceding example consider the discounted process S0,e−r S1,e−2r S2,.... The one step ahead expectations are given byE(e−(n+1)r S n+1|F n)=ue−(n+1)r S n p+de−(n+1)r S n(1−p).The discounted process is a martingale only if the right side is equal to e−nr S n.This is the case only ifp=e r−d u−d.This value of is contained in the unit interval and defines a probability only if d≤e r≤u.In that case the discounted process is a martingale.3.6Martingale RepresentationIn Example3.7we have seen that the process˜S defined by˜S n=e−rn S n in the binomial tree model is a martingale if the tree is equipped with theprobabilityq=e r−d u−d.In this section we shall show that all other martingales in this setting can be derived from˜S in the sense that the increments∆M n=M n−M n−1 of any martingale M0,M1,...must be multiplesφn∆˜S n for a predictable processφ=(φ0,φ1,...).In other words,the change∆M n of an arbitrary martingale at time n−1is proportional to the change in˜S,with the propor-tionality constantφn being a function of the preceding values˜S0,...,˜S n−1 of the process˜S.At time n−1the only randomness to extend M0,...,M n−1 into M n is in the increment∆˜S n.3.8Theorem.If M is a martingle on the binomial tree model of Exam-ple3.1with q=(e r−d)/(u−d)withfiltration F n=σ(S0,...,S n),then there exists a predictable processφ0,φ1,...such that,for every n∈N,∆M n=φn∆˜S n.4Binomial Tree ModelRevisitedSuppose that the price S t at time t is a stochastic process described by the binomial tree model of Example3.1,where it is assumed that the numbers u and d are known.We choose thefiltration equal to F n=σ(S0,...,S n), so that the(only)information available at time n consists of observation of the asset price process until that time.In addition to the asset with price S we can save or borrow money at afixed rate of interest r.We assume that d≤e r≤u.This is a reason-able assumption,because if e r<d then the returns on the asset are with certainty bigger than the return on the savings account,whereas if e r>u, then the returns are with certainty smaller.Then the riskless savings ac-count is never or always preferable over the risky asset,respectively,and a reasonable portfolio will consist of only one type of investment.We equip the branches of the binomial tree with the probability q= (e r−d)/(u−d),rather than a possible real world probability based on historical analysis.Example3.7shows that this gives the unique probability measure on the tree that renders the discounted asset process˜S0,˜S1,..., where˜S n=e−rn S n,into a martingale.A claim is a nonnegative function of C=C(S0,...,S N),where N is the expiry time.Given a claim define the stochastic process˜V=E q(e−rN C|F n).nThe index q on the expectation operator E indicates that we compute ex-pectations under the martingale measure q.In view of Example3.4the process˜V is a martingale.Therefore,by Theorem3.8there exists a pre-dictable processφsuch that(4.1)∆˜V n=φn∆˜S n,n=1,...,N.Given this process we define another processψby(4.2)ψn=˜V n−1−φn˜S n−1.144:Binomial Tree Model RevisitedFrom the facts thatφis predictable and˜V and˜S are adapted,it follows that the processψis predictable.We now interpret(φn,ψn)as a portfolio at time n:(i)φn is the number of assets held during the period(n−1,n].(ii)ψn is the number of units in the saving account during the period (n−1,n].Because both processes are predictable,the portfolio(φn,ψn)can be cre-ated at time n−1based on information gathered up to time n−1,i.e.based on observation of S0,S1,...,S n−1.We shall think of the assets and savings changing value(from S n−1to S n and e r(n−1)to e rn)exactly at time n,and of adapting our portfolio just after time n.Then the value of the portfolio at time n isV n=φn S n+ψn e rn.Just after time n we change the content of the portfolio and the value of the new portfolio is equal toφn+1S n+ψn+1e rn.The following theorem shows that this amount is equal to V n and hence the new portfolio can be formed without additional money:the portfolio process(φ,ψ)is self-financing.Furthermore,the theorem shows that the value V N of the portfolio is exactly equal to the value of the claim C at the expiry time N.As a consequence,we should be indifferent to owning the contract with claim C or the portfolio(φ0,ψ0)at time0,and hence the“just price”of the contract is the value V0of the portfolio.4.3Theorem.The portfolio process(φ,ψ)defined by(4.1)-(4.2)is self-financing.Furthermore,its value process V is nonnegative and satisfies V N=C.Proof.The equation(4.2)that definesψn+1can be rewritten in the form φn+1S n+ψn+1e rn=e rn˜V n,and˜V n=˜V n−1+∆˜V n,where∆˜V n=φn∆˜S n by(4.1).Therefore,φn+1S n+ψn+1e rn−V n=e rn˜V n−V n=e rn˜V n−1+e rnφn∆˜S n−(φn S n+ψn e rn)=e rn˜V n−1+e rnφn(∆˜S n−˜S n)−ψn e rn=e rn˜V n−1−e rnφn˜S n−1−ψn e rn.The right side is zero by the definition ofψn in(4.2).Thus the portfolio is self-financing,as claimed.It also follows from these equations that e rn˜V n−V n=0,whence ˜V=e−rn V n is the discounted value of the portfolio at time n,for every n。

小学下册第6次英语第三单元期中试卷

小学下册第6次英语第三单元期中试卷

小学下册英语第三单元期中试卷英语试题一、综合题(本题有100小题,每小题1分,共100分.每小题不选、错误,均不给分)1. A base feels slippery and can taste ______.2.In a chemical reaction, substances are called __________.3.What is the main ingredient in pancakes?A. FlourB. SugarC. MilkD. Eggs4.The __________ is an important cultural site in Mexico. (玛雅遗址)5.What do you call a garden where flowers grow?A. OrchardB. VineyardC. FlowerbedD. Nursery答案:C6.The ______ (老虎) has sharp claws for hunting.7.The ________ was a significant moment in the history of social change.8. A _____ occurs when the moon is fully illuminated.9.What is the name of the famous wall in China?A. Great Wall of ChinaB. Berlin WallC. Hadrian's WallD. China Wall答案:A10.I saw a rainbow in the ______ (天空) after the rain. It had many ______ (颜色).11.What is 8 divided by 4?A. 1B. 2C. 4D. 6答案:B12.How many days are in February during a leap year?A. 28B. 29C. 30D. 3113.What is the capital of Nicaragua?A. ManaguaB. LeónC. GranadaD. Masaya答案:A14.In autumn, the leaves __________ (变色).15.I enjoy _______ (骑摩托车) on weekends.16.What do we call a young owl?A. OwletB. ChickC. PupD. Calf答案:A17.Lunar rocks brought back by astronauts have been dated to be about ______ billion years old.18.I enjoy watching the ________ dance in the wind.19.The capital of Nigeria is ________.20.I saw a _______ (金鱼) in a bowl.21.I watch _______ after school.22.The __________ is a famous area known for its underwater attractions.23.The __________ is a large body of saltwater that is smaller than an ocean.24.My friend is a _____ (艺术家) who draws comic books.25. A chemical reaction can be represented by a ______ equation.26.The bird is ___ high. (flying)27. A solution that contains the maximum amount of solute is _____ (saturated).28.Matter can exist in different states including _____, liquid, and gas.29.The capital of Belarus is _______.30. A jellyfish has a soft, ______ body.31.What is 18 9?A. 8B. 9C. 10D. 11答案:B32.The _____ (cup) is full of water.33. A ____ is often found swimming in ponds and has smooth skin.34. A __________ (果盘) can show different fruits.35. A ______ is a means of presenting scientific ideas.36.The bumblebee helps pollinate ________________ (花).37.What type of animal is a frog?A. MammalB. ReptileC. AmphibianD. Fish38.She has a pet ________ (狗) at home.39.What is the opposite of ‘true’?A. RealB. FalseC. CorrectD. Right40.I enjoy playing ______ (乐器) like the piano. It makes beautiful ______ (音乐).41.The crow is known for its black ______ (羽毛).42.Which of these is a basic shape?A. CircleB. OctagonC. HexagonD. Rhombus43. A ____ is known for its loud, distinctive call at night.44.The chemical formula for acetic acid is _______.45.What do you call the tool used to measure length?A. ScaleB. RulerC. ThermometerD. Stopwatch答案:B46.What do you call an animal that eats both plants and meat?A. HerbivoreB. CarnivoreC. OmnivoreD. Insectivore答案:C47.What do we call a person who studies the effects of drugs on behavior?A. PsychopharmacologistB. PsychologistC. BiologistD. Chemist答案:A48.The cheetah can run very ______.49.What are the colors of the American flag?A. Red, white, and blueB. Green, yellow, and redC. Black and whiteD. Orange and purple答案:A50.What is the name of the fairy tale character who had long hair?A. CinderellaB. RapunzelC. Snow WhiteD. Belle答案:B51.What is the color of an apple?A. BlueB. GreenC. RedD. All of the above答案:D52.Which fruit is red and often mistaken for a vegetable?A. StrawberryB. TomatoC. CherryD. Raspberry答案:B53. A chemical reaction that produces gas is called a ______ reaction.54.I have a collection of _______ (我有一个_______的收藏).55.I have a toy ________ that can soar through the air.56.The ______ has sharp claws for hunting.57.I like to help my ______ (父母) at home.58.What is the name of the famous artist known for his work with color?A. Claude MonetB. Vincent van GoghC. Pablo PicassoD. Salvador Dalí59.The __________ is the capital city of Brazil. (巴西利亚)60.What is the opposite of hot?A. WarmB. CoolC. ColdD. Burning61.I like to _______ (和家人一起)度假。

总结练习题答案

总结练习题答案

总结练习题答案练习题一:1. What is the capital city of France?Answer: The capital city of France is Paris.2. Who is the author of "Pride and Prejudice"?Answer: The author of "Pride and Prejudice" is Jane Austen.3. What is the chemical symbol for gold?Answer: The chemical symbol for gold is Au.4. Name the main protagonist in the Harry Potter series.Answer: The main protagonist in the Harry Potter series is Harry Potter.5. Who painted the Mona Lisa?Answer: The Mona Lisa was painted by Leonardo da Vinci.练习题二:1. What is the formula for calculating the area of a circle?Answer: The formula for calculating the area of a circle is A = πr², where A represents the area and r represents the radius.2. Who is credited with discovering electricity?Answer: Benjamin Franklin is credited with discovering electricity.3. Which planet is known as the "Red Planet"?Answer: Mars is known as the "Red Planet".4. What is the chemical formula for water?Answer: The chemical formula for water is H2O.5. Who wrote the novel "To Kill a Mockingbird"?Answer: The novel "To Kill a Mockingbird" was written by Harper Lee.练习题三:1. What is the formula for calculating the perimeter of a rectangle?Answer: The formula for calculating the perimeter of a rectangle is P = 2(l + w), where P represents the perimeter, l represents the length, and w represents the width.2. Who is the current President of the United States?Answer: The current President of the United States is Joe Biden.3. What is the symbol for the chemical element oxygen?Answer: The symbol for the chemical element oxygen is O.4. Who is the author of the book "1984"?Answer: The author of the book "1984" is George Orwell.5. Name the largest ocean in the world.Answer: The largest ocean in the world is the Pacific Ocean.练习题四:1. What is the formula for calculating the volume of a cylinder?Answer: The formula for calculating the volume of a cylinder is V = πr²h, where V represents the volume, r represents the radius of the base, and h represents the height.2. Who is the founder of Microsoft?Answer: The founder of Microsoft is Bill Gates.3. Which gas makes up the majority of Earth's atmosphere?Answer: The gas that makes up the majority of Earth's atmosphere is nitrogen.4. Who wrote the play "Romeo and Juliet"?Answer: The play "Romeo and Juliet" was written by William Shakespeare.5. What is the largest continent in the world?Answer: The largest continent in the world is Asia.练习题五:1. What is the formula for calculating the area of a triangle?Answer: The formula for calculating the area of a triangle is A = 1/2bh, where A represents the area, b represents the base, and h represents the height.2. Who is the author of "The Great Gatsby"?Answer: The author of "The Great Gatsby" is F. Scott Fitzgerald.3. What is the chemical symbol for iron?Answer: The chemical symbol for iron is Fe.4. Who is the main character in the novel "The Catcher in the Rye"?Answer: The main character in the novel "The Catcher in the Rye" is Holden Caulfield.5. Name the largest lake in Africa.Answer: The largest lake in Africa is Lake Victoria.。

Discriminant Analysis – Basic Relationships

Discriminant Analysis – Basic Relationships


In some analyses, we might discover that two or more of the groups defined by the dependent variable cannot be distinguished using the available independent variables. While it is reasonable to interpret a solution in which there are fewer significant discriminant functions than the maximum number possible, our problems will require that all of the possible discriminant functions be significant.

SW388R7 Data Analysis & Computers II Slide 8
Groups, functions, and variables

To interpret the relationship between an independent variable and the dependent variable, we must first identify how the discriminant functions separate the groups, and then the role of the independent variable is for each function.

This interpretation is complicated by the fact that the relationship is not direct, but operates through the discriminant function. Dependent variable groups are distinguished by scores on discriminant functions, not on values of independent variables. The scores on functions are based on the values of the independent variables that are multiplied by the function coefficients.

麦田里的守望者英文书评

麦田里的守望者英文书评

Major ThemesPainful Exp erie nee vs. Numb nessP erha ps the greatest theme of the no vel invo Ives the relati onship betwee n the pain of actual exp erie nee and feeli ng on e's feeli ngs, on the one hand, and on the other hand the equally devastating numbness that comes with shutting dow n on e's emoti ons in order to avoid sufferi ng. After the death of Allie, Holde n esse ntially shuts dow n, forcing himself to lose all attaehme nts to people so as n ever to be hurt aga in. He rep eatedly men ti ons how imp orta nt it is not to get attached to anyone, since this will lead to miss ing them once they are gone. By the end of the no vel, he has sp iraled so far dow n with this theory that he has become afraid to eve n sp eak to anyone. P hoebeis p erha ps the only rem in der that Holde n still has the cap acity to love. Whe n he looks at her, he cannot help but feel the same tortured love that he felt for Allie. Nevertheless, the surges of these feeli ngs leave him eve n more bereft. He knows he must leave Phoebe to protect himself, but when she shows up to accompany him on his journey, ultimately he puts his love for her first and sacrifices his own in st inct to flee in order to retur n home.Holde n, it seems, is in the throes of an existe ntial crisis. To a great degree he is numb to the pains and joys of life. Un able to come to terms with his brother's death, he has no one to show him the kind of paren tal or brotherly love that he himself gave Allie. Whe never some one does end up show ing him eve n a hint of such love (such as Mr. An toli ni), Holde n ends up being disa ppoin ted.Love and SexAt his core, Holde n is a dee p, sen sitive soul, at bottom un able to sublimate hisfeeli ngs into numbn ess. He env ies some one like Stradlater, who can simp ly pick up girls whe never he likes, and who treats sex as a casual p leasure. To Holde n, however, sex is dee ply discomfort ing. He cannot have it with girls he likes, and he cannot man age to numb himself eno ugh to treat girls casually. Numbi ng himself to love, it seems, is Holde n's greatest challe nge. He feels too dee ply about the world, about people, to truly shut dow n. Whe n he fin ally does fall in love with Jane Gallagher, he soon discovers that Stradlater has a date with her, which con firms his sus picion that everythi ng he loves eve ntually deteriorates. He leaves Pen cey with some hope of inven ti ng a new ide ntity, but he cannot break out of his being. Even in the presenee of a prostitute, he cannot thi nk of hav ing sex, only of hav ing a conv ersatio n in the hope of feeli ng some glimmer of huma n affecti on with her. All Holde n wants to do is talk, but he cannot find some one who will liste n.Loss of InnocenceHolde n must face that fork in the road of adolesce nee whe n one realizes that maturity en tails a loss of innocence — that greater kno wledge of on eself and others and the circumstances all comes with a price. In Holden's case, he cannot bear to acce pt the death of Allie, the death of pure innocence that had no good reas on to suffer or die. I n Holde n's eyes, Allie is truth, while every one else is phony. ” Innocence goes with idealism and a certa in in ability or unwillingness to bear and accept the harsherreality. Holden cannot bear to hold onto his innocence because innocence brings its own harms; people con ti nue to disa ppoint him. Thus the cost of maturity is much less; innocence has been quite painful, too.Innocence dema nds more money for nothing, thepedophile, and the cab drivers beratequestions about the birds in the park.p reserve his innocen ce, this is not eno ugh,for he cannot find real love in the outsideworld. Besides, losing Allie has brought tremendous pain. Holden also has the com mon adolesce nt exp erie nee of p ercei ving that time in school lear ning mundane less ons feels p etty whe n his en tire soul is in flux as it comes to grips with reality. When the en tire world around him app ears phony, where can he go to grasp hold of some reality, some stable truth? Without an explanation why Allie was taken from him, there appears no reason behind the world's events, and in this respect Holden's maturity invoIves a deep loss of innocence such that he perceives that the reality of the world is its irrati on ality. Phonin ess vs. Authe nticityHolde n labels almost every one a pho ny, ” exce pti ng P hoebe, Allie, himself. In Holde n's eyes, a pho ny ” is some one who embraces the world mundane dema nds and tries to make someth ing out of nothing — that is, just about every one who studies in school or who puts on airs in order to do a job or achieve a goal. The fact that no one is ack no wledg ing how trivial and fleeting life is, compared with the grand things we tell one another about reality — how difficult it is to truly love and share on eself with people knowing that all, like Allie, will eve ntually die — causes him to bur n with frustrati on, eve n rage. Holde n un dersta nds on some level one of the most profound truths of mortal life: the sup erficial matters little because it will not last, yet it is made to seem so much more important. Meanwhile, all around him, he must watch has been problematic: the prostitute man who takes him in seems like a him as stupid when he asks simple While Allie 's memory can help him very and ssup erficial people win honors through their artifice. He thus holds his dee pest contempt for those who succeed as phoni es: Stradlater, the Headmaster, and all the boys who treat school as if it is a club to be ruled by Social Darwi nism. All Holde n wan ts is some authe ntic livi ng, to hold on to some one like P hoebe or Allie who knows nothing of the world ' superficiality and therefore is not tain ted by it, but he is afraid to make it too real out of the justified fear of one day los ing them forever.Life and DeathA key part of Holde n ' emoti onal life invo Ives his react ion to Allie ' death. People live for a while, but all too soon we all die. Allie did not choose it, but Holde n thi nks about James Castle, a skinny boy who jumped out the win dow at school and fell to his death. Holde n himself en terta ins thoughts of a similar suicide. The decisi onto numb himself to his feeli ngs about life is a decisi on to shut himself dow n emoti on ally so much that he is no Ion ger truly livi ng. It is a decisi on, however, that remai ns fun dame ntally impo ssible for Holde n. When he thinks about James Castle, he cannot bear to imagine James just laying there amidst the stone and blood, with no one p ick ing him up.Holde n might see some roma nee in suicide and some comfort in the idea that it ends internal pain, but death does seem worse, the ultimate Ion eli ness. He see n the effects of death on the liv ing as well. He thus cannot do to P hoebe what Allie has done to them already.He pl ods on, only sure that he must gradually wea n himself away from P hoebe so that she gets used to los ing him forever--a nd so that he gets used to being away from her. Though Holde n n eeds close ness and love in order to renew his life, he kee ps driv ing himself further away from it in order to avoid the in evitable loss. The more he wants to exp erie nee life, the more an tisocial he becomes and the more he imagi nes death. This p aradox is part of Holde n ' life: there is pain in shutt ing dow n on e's feeli ngs, and there is pain in the risk of opening on eself up aga in. He impo ssibly tries to avoid pains that are in evitable for huma n mortals while they live.Lack of Authority FiguresHolde n is profoun dly alone. His parents are abse nt exce pt for in sist ing that he progress along a conventional path and stay in school as long as he can before he is kicked out or tires of each in stituti on. His parents do not let him regro up but send him off to the n ext school. At Pen cey, Holde n finds no adult totrust with his feeli ngs; most people everywhere are phony. Some adults eve n seem so selfish that they are willi ng to abuse childre n. Overall, Holde n views adults with intense disa ppoin tme nt, eve n cyni cism. How is it that the older they get, thefarther from authenticity they get? Meanwhile, the gradual deteriorati on of the body disgusts him. Upon visit ing an old pro fessor, much of his thoughts are dedicated to the awfuln ess of the old man's body. There is no allure in grow ing older.Authority does not seem related to wisdom, either. Adults tell Holden to find directi on and thus stability, but he views such advice as both sus picious and n a?ze; p lay ing such a game is in authe ntic. Going his own way aut onom ously, as a law unto himself, does not work out so well either, so it is un clear where Holde n might find legitimate authority.Lon eli nessHolde n is very Ion ely, and his adolesce nt Ion eli ness seems to run much dee per tha n the feeli ngs so com monly felt at that age. He admits to his Ion eli ness openly, and it gives him evide nee that p erha ps he might still have some emotio ns left. At the same time, Holde n takes few ste ps to mitigate his Ion eli ness. Whe never he feels the urge to meet some one, to call up a girl, to have a social exp erie nee, he ends up sabotag ing it before he can get hurt. He thus protects himself so fully that he effectively shuts off any possibilities of alleviat ing his own Ion eli ness. He might want to call Jane, for exa mp le, but he hangs up before she gets on the phone. He might want to sleep with a p rostitute to feel huma n comfort, but this will not do. He might want to in teract with friends at a bar, but he ends up saying something hurtful so that they aba ndon him. Pushing them away pro vides a dee per and dee per Ion eli ness, but at these mome nts of choice he is willi ng to en dure it rather tha n eve ntually face the ultimate, devastat ing Ion eli ness of losi ng ano ther person like Allie出师表两汉:诸葛亮先帝创业未半而中道崩殂,今天下三分,益州疲弊,此诚危急存亡之秋也。

  1. 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. 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. 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)。
arXiv:math/0509117v2 [math.DS] 15 Jun 2007
A trace formula for the forcing relation of braids
Boju Jiang∗, Hao Zheng†
Department of Mathematics, Peking University Beijing 10871, China2
{(ht (xi ), t) | 0 ≤ t ≤ 1, 1 ≤ i ≤ n} in the cylinder R2 × [0, 1]. Indeed, the closed curve {[ht (x1 ), . . . , ht (xn )] | 0 ≤ t ≤ 1} in the configuration space Xn = {(x1 , . . . , xn ) | xi ∈ R2 , xi = xj , ∀i = j }/Σn , where Σn denotes the symmetric group of n symbols, gives rise to a braid βP in the n-strand braid group Bn = π1 (Xn ). With another connecting isotopy {ht }, the resulting braid βP may differ by a power of the “full-twist”. Matsuoka obtained lower bounds for the number of m-periodic points of f |R2 \P , in terms of the trace of the reduced Burau representation of the braid (βP )m . Later, Kolev [20] (see also [12]) found that a 3-periodic orbit P guarantees the existence of m-periodic orbits for every m, unless the braid βP is conjugate to a power of the braid σ1 σ2 . Roughly speaking, this means that the (2π/3)-rotation mentioned above is the only exceptional case. Therefore, Li-Yorke’s statement still holds in a subtle way in 2-dimensional dynamics. The analogue of the Sharkovskii order naturally leads to the notion of forcing relation of (conjugacy classes of) braids. In the following, the notation [β ] stands for the conjugacy class (in the group which is specified by the context) of a braid β . Definition 1.2. A braid β forces a braid γ if, for any orientation-preserving homeomorphism f : R2 → R2 and any isotopy {ht } : id ≃ f , the existence of an f -invariant set P with [βP ] = [β ] guarantees the existence of an f -invariant set Q with [βQ ] = [γ ]. Remark 1.3. There is a homomorphism from the braid group Bn onto the mapping class group of the pair (R2 , P ) (acting on (R2 , P ) from the right ), its kernel being generated by the “full-twist”. Via this homomorphism, [βP ] is sent to the conjugacy class of the mapping class represented by f , which is independent of the choice of the isotopy {ht }. Following Boyland [5] this invariant is referred to as the braid 2
type of (f, P ) in the literature. It is clear that the forcing relation of braids defined above naturally descends to that of braid types. The forcing relation is essentially a problem concerning plane homeomorphisms. So the Bestvina-Handel theory of train-track maps [2] comes in naturally. By analyzing the symbolic dynamics of train-track maps, Handel [14] was able to totally solve the forcing relation among 3-strand pseudo-Anosov braids, and de Carvalho and Hall [7, 8] have managed to do the same for horseshoe braids. This approach is, theoretical speaking, powerful enough to be extended to mapping classes of all punctured surfaces. On the other hand, there still is the challenging task of recovering the braiding information encoded in the symbolic dynamics. In this paper, we take another approach. Besides the Thurston classification of surface homeomorphisms, we apply the Nielsen fixed point theory. As a powerful tool for studying fixed points and periodic orbits of self maps, Nielsen theory has been well developed and successful in many mathematical problems. It turns out that there are plenty of coincidences between the notions in the Nielsen theory and the forcing theory, providing a more direct bridge between the topological and the algebraic aspects of braids. We start by slightly expanding the language of forcing. Definition 1.4. A braid β ′ is an extension of β if β ′ is a (disjoint but possibly intertwined) union of β and another braid γ . An extension β ′ is forced by β if, for any orientation-preserving homeomorphism f : R2 → R2 and any isotopy {ht } : id ≃ f , the existence of an f -invariant set P with [βP ] = [β ] guarantees the existence of an additional f -invariant set Q ⊂ R2 \ P with [βP ∪Q ] = [β ′ ]. The advantage of considering [βP ∪Q ] is that it contains the extra information of how the forced braid γ = βQ winds around the original braid β . Our main result is stated as follows. Theorem 1.5. Suppose a braid β ′ ∈ Bn+m is an extension of β ∈ Bn . Then β ′ is forced by β if and only if β ′ is neither collapsible nor peripheral relative to β , and the conjugacy class [β ′ ] has nonzero coefficient in trBn+m ζn,m(β ). In the theorem, ζn,m is a matrix representation of Bn over a free ZBn+m -module, and the trace trBn+m is meant to take value in the free abelian group generated by the conjugacy classes in Bn+m (see Section 4). In addition, β ′ is said to be collapsible or peripheral relative to β if, roughly speaking, some strands of β ′ may be merged or moved to infinity while keeping β untouched (see Definition 3.3 and the figures therein). Thus, to obtain the (n + m)-strand forced extensions of a braid β ∈ Bn , it suffices to compute the trace trBn+m ζn,m(β ) and then drop off certain irrelevant terms. 3
相关文档
最新文档