Book Description
A comprehensive look at learning and generalization theory. The statistical theory of learning and generalization concerns the problem of choosing desired functions on the basis of empirical data. Highly applicable to a variety of computer science and robotics fields, this book offers lucid coverage of the theory as a whole. Presenting a method for determining the necessary and sufficient conditions for consistency of learning process, the author covers function estimates from small data pools, applying these estimations to real-life problems, and much more.
Customer Reviews:
An excellent overview.......2004-07-22
The field of statistical learning theory has not only seen considerable advances in the last fifteen years, it has also found many applications, some of these appearing in commercial packages. It is now classified as a subfield of artificial intelligence, and as such gives an alternative, and frequently more general viewpoint on such topics as pattern recognition, regression estimation, and signal processing. The author of this book is one of the originators of statistical learning theory, and has written a book that will give the mathematically sophisticated reader a rigorous account of the subject. Most of the main results are proven in detail, but the author does find time to include insightful discussion on the origins and intuition behind the concepts involved in statistical learning theory.
Along with a brief introduction, the book consists of three parts, the first being an overview of the statistical theory of learning, the second giving the details of the now widely used support vector machines, and the last one (the most sophisticated mathematically) giving the statistical foundations of learning theory. In writing the book, the author wants to put forward a new approach to dependency estimation problems having their origin in learning theory, and being able to deal with the ?curse of dimensionality?. The origins of the subject lie in the pattern recognition problem and the Glivenko-Cantelli problem in statistics. Both of these problems were discovered to be essentially the same, and the author?s task is to use their similarities to construct a general theory of statistical inference and (inductive) learning. Indeed, a new induction principle, called ?structural risk minimization? (SRM) is paradigmatic in the book, along with the now ubiquitous VC dimension, the latter of which originates in the author?s early research. Both the SRM and the VC dimension illustrate the tension between the need for high accuracy and the need for the minimization of error in data sets.
The learning problem, as the author sees it, is the problem of selecting the correct dependence on the basis of empirical data. Two approaches to this problem are discussed, the first using a ?risk functional?, and the second involving the estimation of stochastic dependencies and the consequent solution of integral solutions. Both of these approaches are modeled in terms of a general model of learning from examples, which consists of a data generator, a supervisor, and a learning machine. The learning machine can either imitate the supervisor or identify how the supervisor operates. These two methods are different, the author says, in that the first one searches for the best prediction based on the data, while the second one attempts to approximate the operator representing the supervisor. Both approaches are studied in the book, with the first one being the easier of the two, while the second involving the solution of ill-posed problems. The author views the learning process in terms of choosing the right function from a given function collection.
Both perceptrons and their generalizations, neural networks, are briefly discussed in the book, along with the back-propagation method. The author gives reasons why he does not think neural networks are well-controlled learning machines, such as the existence of local minima, the slow convergence of the gradient method, and the choice of scaling factors. These problems serve as motivation for the introduction of support vector machines, which are introduced as optimal separating hyperplanes. Support vector machines take input vectors into a high-dimensional feature space via a nonlinear mapping, and an optimal separating hyperplane is then constructed in this feature space.
Similar to the need for neural networks to generalize well, separating hyperplanes must do the same, and due to the large dimensionality of the feature space, a hyperplane that separates the training data may not generalize well. In addition, the large dimensionality of the feature space makes the construction of the hyperplane computationally demanding. The author shows that optimal hyperplanes, found using various mathematical techniques such as quadratic optimization, do generalize well. Also, as the author points out, the explicit form of the feature space need not be known, since only the inner products between the ?support vectors? and the vectors of the feature space need to be calculated. The calculation of the inner product is done with the insight gained from Mercer?s theorem, which gives the existence of a kernel function such that there exists a feature space where this function generates the inner product. This inner product in feature space allows the construction of a decision function that is nonlinear in the input space but that is equivalent to a linear function in the feature space. Different choices of the kernel function give different types of learning machines. The author discusses three examples of support vector machines for pattern recognition: polynomial, radial basis function, and two-layer neural network support vector machines. An entire chapter is spent on the problem of digit recognition using support vector machines.
new approach to inference based on VC dimension.......2002-01-04
Vapnik and Chernovenkis extended the Glivenko-Cantelli Theorem in their work on classification and statistical learning. Vapnik in recent texts has described a form of nonparametric statistical inference based on approximating functions and the Vapnik-Chernovenkis dimension.
In an earlier book published by Springer-Verlag he develops the basics of the theory. However to keep the mathematical level excessible to computer scientists and engineers he avoided the mathematical proofs needed for mathematical rigor. This text is an advanced text that provides the rigorous development. Although the preface and chapter 0 give the reader a idea of what is to come the rest of the text is difficult reading.
The theory has been quite successful at attacking the pattern recognition/ classification problem and provides a basis for understanding support vector machines. However Vapnik sees a much broader application to statistical inference in general when the classical parametric approach fails.
If you have a strong background in probability theory you should be able to wade through the book and get something out of it. If not I recommend reading section 7.9 of "The Elements of Statistical Learning" by Hastie, Tibshirani and Friedman. That will give you an easily understandable view of the VC dimension. Also sections 12.2 and 12.3 of their text will give you some appreciation for support vector machines and the error rate bounds obtainable for them based on the VC dimension.
Rich & Valuable.......2001-07-25
This book aims at rigorours and deep treatment of statistical learning and is divided into three parts :
(I)THEORY OF LEARNING AND GENERALIZATION;
(II)SUPPORT VECTOR ESTIMATION OF FUNCTIONS;
(III)STATISTICAL FOUNDATION OF LEARNING THEORY'
For anyone intending to dive into this topic intriguing readers shull find their task rather not simple when exploring this mathematical exposition.This is because of the mature nature behind the basic theory .In order to gain most of the benefit ,interested and even involved researchers are urged and should assume all the requirements for a vast and solid mathematical background.
I Think the book constitutes a respectful and organized 'exhibition' that you will not find in any other place. Althought there are excellent books discussing SVMs and Machine-Learning/ Intelligence,eventually all emenate from the theory.Regarding the book rating it is was not rated upon how much you retrieve as concepts, but how well the propositions offer a precious appreciation of the substantial theory.In otherwords, this book is not the place for a first time learning, but it is serves as a bridge between interrelated elements of such incredibly growing area.
For the book: "The Nature of Statistical learning Theory" also by Vapnik you can find a review by Vladimir Cherkassky in The IEEE TRANSACTIONS ON NEURAL NETWORKS VOL. 8, NO. 6, NOVEMBER 1997 .
Book Description
A crucial challenge for economists is figuring out how people interpret the world and form expectations that will likely influence their economic activity. Inflation, asset prices, exchange rates, investment, and consumption are just some of the economic variables that are largely explained by expectations. Here George Evans and Seppo Honkapohja bring new explanatory power to a variety of expectation formation models by focusing on the learning factor. Whereas the rational expectations paradigm offers the prevailing method to determining expectations, it assumes very theoretical knowledge on the part of economic actors. Evans and Honkapohja contribute to a growing body of research positing that households and firms learn by making forecasts using observed data, updating their forecast rules over time in response to errors. This book is the first systematic development of the new statistical learning approach.
Depending on the particular economic structure, the economy may converge to a standard rational-expectations or a "rational bubble" solution, or exhibit persistent learning dynamics. The learning approach also provides tools to assess the importance of new models with expectational indeterminacy, in which expectations are an independent cause of macroeconomic fluctuations. Moreover, learning dynamics provide a theory for the evolution of expectations and selection between alternative equilibria, with implications for business cycles, asset price volatility, and policy. This book provides an authoritative treatment of this emerging field, developing the analytical techniques in detail and using them to synthesize and extend existing research.
Book Description
The aim of this book is to discuss the fundamental ideas which lie behind the statistical theory of learning and generalization. It considers learning as a general problem of function estimation based on empirical data. Omitting proofs and technical details, the author concentrates on discussing the main results of learning theory and their connections to fundamental problems in statistics. These include: * the setting of learning problems based on the model of minimizing the risk functional from empirical data * a comprehensive analysis of the empirical risk minimization principle including necessary and sufficient conditions for its consistency * non-asymptotic bounds for the risk achieved using the empirical risk minimization principle * principles for controlling the generalization ability of learning machines using small sample sizes based on these bounds * the Support Vector methods that control the generalization ability when estimating function using small sample size. The second edition of the book contains three new chapters devoted to further development of the learning theory and SVM techniques. These include: * the theory of direct method of learning based on solving multidimensional integral equations for density, conditional probability, and conditional density estimation * a new inductive principle of learning. Written in a readable and concise style, the book is intended for statisticians, mathematicians, physicists, and computer scientists. Vladimir N. Vapnik is Technology Leader AT&T Labs-Research and Professor of London University. He is one of the founders of statistical learning theory, and the author of seven books published in English, Russian, German, and Chinese.
Customer Reviews:
New to Field of Learning Theory.......2006-04-11
I am relatively new to statistical learning theory, though with a solid background in supporting theories and a Master's in Engineering. I found the text readable. I appreciate the historical perspective and the development of concepts by the author. I was generally able to grasp Vapnick's theories and explanations, though often after rereading passages many times.
Simple examples would significantly aid the readability and understandability of the text - akin to the way we teach our children. We don't describe all the attributes of a rabbit, we point to a picture of a rabbit and say "bunny". After two or three examples of this my children know the abstract concept of a rabbit (without me having to describe a small, four legged creature with long ears, etc. and then answering the inevitable question of "What's four legged creature mean daddy?"). Particularly with a text about learning theory, one would think it would be full of such examples - at least from a pedagogical point of view.
Initially, I didn't mind Vapnick's editorializing, but after a while I find it annoying - I'm sure he didn't single-handedly invent the entire field of statistical learning theory, but he sure doesn't miss any opportunities to tell the reader that he believes he has.
worth reading.......2001-09-22
A good, albeit highly idiosyncratic, guide to Statistical Learning. The highly personal account of the theory is both the strong point and the drawback of the treatise. On one side, Vapnick never loses sight of the big picture, and gives illuminating insights and formulations of the "basic problems" (as he calls them), that are not found in any other book. The lack of proofs and the slightly erratic organization of the topic make for a brisk, enjoyable reading. On the minus side, the choice of the topics is very biased. In this respect, the book is a self-congratulatory tribute by the author to himself: it appears that the foundations of statistical learning were single-handedly laid by him and his collaborators. This is not really the case. Consistency of the Empircal Risk Measure is rather trivial from the viewpoint of a personal trained in asymptotic statistics, and interval estimators for finite data sets are the subject of much advanced statistical literature. Finally, SVMs and neural nets are just a part of the story, and probably not the most interesting.
In a nutshell, what Vapnick shows, he shows very well, and is able to provide the "why" of things as no one else. What he doesn't show... you'll have to find somewhere else (the recent Book of Friedman Hastie & Tibs is an excellent starting point).
A last remark. The book is rich in grammatical errors and typos. They could have been corrected in the second edition, but do not detract from the book's readability.
A very nice book to get ideas on support vector machines.......2000-08-28
This is a very readable book by an authority on this subject. The book starts with the statistical learning theory, pioneered by the author and co-worker's work, and gradually leads to the path of discovery of support vector machines. An excellent and distinctive property of support vector machines is that they are robust to small data perturbation and have good generalization ability with function complexity being controlled by VC dimension. The treatment of nonlinear kernel classification and regression is given for the first time in the first edition. The 2nd edition includes significant updates including a separate chapter on support vector regression as well as a section on logistic regression using the support vector approach. Most computations involved in this book can be implemented using a quadratic programming package. The connections of support vector machines to traditional statistical modeling such as kernel density and regression and model selection are also discussed. Thus, this book will be an excellent starting point for learning support vector machines.
A research field described by the man who invented it.......2000-02-25
Vapnik and collaborators have developed the field of statistical learning theory underlying recent advances in machine learning and artificial intelligence (e.g. support vector machines). This book almost accomplishes the formidable task of comprehensibly describing the essential ideas of learning theory to non-statisticians. It contains ample theorems but almost no proofs.
Book Description
With Bayesian network technology very much on the up-swing in industry and government, there is an increasing need for an introductory book that jointly emphasizes the understanding of its underlying priniciples and their application in practice. Bayesian Artificial Intelligence presents elements of Bayesian network technology, automated causal discovery, and learning probabilities from data along with extensive motivational examples of using these technologies to develop probabilistic expert systems. This practical, very accessible introduction balances the causal discovery of networks with the Bayesian inference procedures that use a network once it is found. The authors emphasize understanding and intuition, so they keep the mathematical details to a minimum, but also provide the algorithms and technical background needed for applications. They illustrate at length a number of applications and discuss application software in detail. A broad range of topics, a practical perspective, and a thoughtful discussion of philosophical underpinnings make Bayesian Artificial Intelligence an ideal introduction for students and for professionals who want to broaden their expertise. It provides the knowledge you need to put Bayesian network tools into practice, while also forming the basis for a more detailed investigation of the technology and original research.
Customer Reviews:
Very good introduction in causal Modeling.......2006-03-09
The book by Korb and Nicholson is very readable and structured. Starting with some background information in statistics it comes directly to the major topic of the book - bayesian networks. The theory thereof is nicely evolved and applied to small examples to demonstrate its usage. Each chapter finishes with a short summary and bibliographical notes for further readings.
In my opinion this book is well written and the chosen examples are insightful. What I do not like is part three of the book which is devoted to case studies and praktical examples. If this space had been used for the first two parts by providing more details, e.g., for the discussion of path models (which is given but only short), this book could be even great on a more advanced level. In this form it is very good as an introduction in Bayesian Networks and related topics like the larger class of causal models.
Excellent Introductory Text.......2004-12-17
It is difficult to assess a review without understanding the biases of the reviewer. I fall under the category of researcher/practitioner when it comes to reasoning with graphical models. I am familiar with and make use of several books and papers on this topic in my work. Of the set of standard references (Pearl, Jensen, Neapolitan, Jordan, Cowell et al., Borgelt & Kruse) the text by Korb and Nicholson (K&N) stands out in terms of its clarity and accessibility. Does the book have everything one would ever want to know about Bayesian inference? Not by a long shot. Is it, however, a good place to start? Definitely. The basic concepts are presented relatively completely and with clarity. I consistently recommend K&N over other alternatives to colleagues new to the field. Is there a chasm separating concept and algorithm in the book? I don't think there is, especially relative to other references. With tools such as Kevin Murphy's BNT, or Netica available on the Web, it seems to me that providing a solid conceptual framework becomes paramount for a textbook such as this. I believe K&N succeed admirably in this sense. Why four stars and not five? Even for an introductory text such as K&N, it would be nice to have more development of some concepts such as causality, context specific independence, or loss of independence in dynamic nets. Although it won't be your last book on reasoning with graphical models, K&N should probably be your first.
Bayesian Networks for Undergrads and Practicioners.......2004-01-12
Despite its name "Bayesian Artificial Intelligence" covers Bayesian network (BN) techniques only. Other Bayesian techniques useful for AI are not treated.
The content is divided in three main sections: (1) The basics of probabilistic reasoning with BNs, (2) Causal discovery (finding BNs from data), and (3) "Knowledge engineering".
The first part covers the fundamental concepts and algorithms around BNs and (simple) decision networks. It is well-written and clear, but readers who are not totally new to the field might find only little new information (e.g., loopy belief propagation, continuous densities, large decision networks, etc. are not covered).
The second part is on how to deduce causal relationships from observational data. Constrained-based and Bayesian approaches are covered, but on a rather general level. I am not sure how easy it is to implement the algorithms from the descriptions provided. When it comes to details of the algorithms, proofs, or mathematical background the authors very often refer to the literature due to "lack of space". From a practical standpoint, it is unfortunate that the different methods are compared to each other only superfiscially. For instance, one method presented performs a large number of statistical tests; one would expect that this requires large amounts of data in order to avoid false positive results. Is this a problem? With questions like these the reader is often left alone.
I am not competent to talk about part three (knowledge engineering), so I end with my general impression of the book: I would have appreciated if the authors had treated some the algorithms in greater detail and had spent a few pages on advanced concepts and current research directions. On the other hand, some information provided could have easily been left out. (For instance, how to download and install certain software packages from the internet, Kevin Murphy's well-known survey on BN software packages, screenshots of user dialogs, etc. just eat pages. Providing the URLs to the corresponding sites on the internet is completely sufficient, and the information there is more likely to be up-to-date.) The saved pages could then be spent on information which is not readily available elsewhere.
To summarize: The book provides a mostly well-written general overview of the basic concepts and could serve as a first introduction to the field. However, it leaves the reader often alone when it comes to the mathematical background, potential practical pittfalls, or advanced algorithms.
Book Description
The minimum description length (MDL) principle is a powerful method of inductive inference, the basis of statistical modeling, pattern recognition, and machine learning. It holds that the best explanation, given a limited set of observed data, is the one that permits the greatest compression of the data. MDL methods are particularly well-suited for dealing with model selection, prediction, and estimation problems in situations where the models under consideration can be arbitrarily complex, and overfitting the data is a serious concern.
This extensive, step-by-step introduction to the MDL Principle provides a comprehensive reference (with an emphasis on conceptual issues) that is accessible to graduate students and researchers in statistics, pattern classification, machine learning, and data mining, to philosophers interested in the foundations of statistics, and to researchers in other applied sciences that involve model selection, including biology, econometrics, and experimental psychology. Part I provides a basic introduction to MDL and an overview of the concepts in statistics and information theory needed to understand MDL. Part II treats universal coding, the information-theoretic notion on which MDL is built, and part III gives a formal treatment of MDL theory as a theory of inductive inference based on universal coding. Part IV provides a comprehensive overview of the statistical theory of exponential families with an emphasis on their information-theoretic properties. The text includes a number of summaries, paragraphs offering the reader a "fast track" through the material, and boxes highlighting the most important concepts.
Customer Reviews:
VERY heavy on theory & math.......2007-09-01
This book is provides a good overview to the theory behind MDL. Not for the faint of heart, however.
Book Description
This text focuses on the skills and processes necessary for understanding statistical research in language learning Designed for language teachers with no previous background in statistics, the paperback edition focuses on the skills and processes necessary for understanding statistical research in language learning. Brown explains the basic terms of statistics, the structure and organization of statistical research reports, the system of statistical logic, and how to decipher tables, charts, and graphs. By the end of the book, readers will be able to make knowledgeable judgments about the relative qualities of a study and to assess the value of the results of a study in relation to a specific language teaching situation.
Customer Reviews:
research design and statistics for the language teacher.......2000-08-14
"Understanding Research in Second Language Learning" is the perfect book for any teacher or researcher who would like to be able to critique or assess statistical research papers without having to spend an inordinate amount of time studying math in the process. The book starts off by asking the pertinent question "What is research?", before explaining clearly and concisely what variables are and how they can be identified, how data is measured, and how to choose the most appropriate statistical analysis for your particular research question. What I liked most about this book was the methodological approach to understanding what good research design is, and that the helpful diagrams, summaries and review questions made understanding rather easier than may have been otherwise. With this book, you will probably be able to understand ninety percent of the stats that appear in language research papers, and to follow the logic of almost all.
Book Description
In Reliable Reasoning, Gilbert Harman and Sanjeev Kulkarni--a philosopher and an engineer--argue that philosophy and cognitive science can benefit from statistical learning theory (SLT), the theory that lies behind recent advances in machine learning. The philosophical problem of induction, for example, is in part about the reliability of inductive reasoning, where the reliability of a method is measured by its statistically expected percentage of errors--a central topic in SLT.
After discussing philosophical attempts to evade the problem of induction, Harman and Kulkarni provide an admirably clear account of the basic framework of SLT and its implications for inductive reasoning. They explain the Vapnik-Chervonenkis (VC) dimension of a set of hypotheses and distinguish two kinds of inductive reasoning, describing fundamental results about the power and limits of those methods in terms of the VC-dimension of the hypotheses being considered. The VC-dimension is found to be superior to a related measure proposed by Karl Popper, and shown not to correspond exactly to ordinary notions of simplicity. The authors discuss various topics in machine learning, including nearest-neighbor methods, neural networks, and support vector machines. Finally, they describe transductive reasoning and suggest possible new models of human reasoning suggested by developments in SLT.
Customer Reviews:
A great little book.......2007-06-06
I had the great priviledge of taking the class upon which this book was based last semester at Princeton University under professors Harman and Kulkarni. It is a fascinating little book, which manages to distill decades of debate and research into concise, readable chapters that carry the presentation forward. The authors' approach is original but commonsensical and they clearly demonstrate the value of interdisciplinary work in their twin fields of philosophy and electrical engineering!
The book is not without its flaws, however. The first chapter seems to take off 'in medias res' expecting the reader to be fully caught up with the latest discussion on the problem of induction, and it is not always clear exactly what a 'process of reasoning' might be compared to deductive arguments. The discussion could have benefited from incorporating material from the other draft textbook we used in class, on "The Nature and Limits of Learning", and even from the lecture handouts. The discussion of simplicity, as well, could have been clarified, especially with regard to Goodman's new riddle of induction and Karl Popper's philosophy of science.
Also rather disappointing in class was the discovery that Harman and Kulkarni's method do not warrant going beyond instrumentalism in scientific theorizing. I was hoping for something a little more robust. In any case, this book should be read by anyone interested in the issues they raise. It sure got me thinking and I will definitely refer to it later on as my research in philosophy brings me in contact again with the issues they discuss.
Average customer rating:
|
Statistical Mechanics Of Learning
Andreas Engel
Manufacturer: Cambridge University Press
ProductGroup: Book
Binding: Paperback
 General
| Artificial Intelligence
| Computer Science
| Computers & Internet
| Subjects
| Books
Neural Networks
| Artificial Intelligence
| Computer Science
| Computers & Internet
| Subjects
| Books
Theory of Computing
| Artificial Intelligence
| Computer Science
| Computers & Internet
| Subjects
| Books
Computer Mathematics
| Artificial Intelligence
| Computer Science
| Computers & Internet
| Subjects
| Books
General
| Computers & Internet
| Subjects
| Books
General
| Education
| Nonfiction
| Subjects
| Books
Probability & Statistics
| Applied
| Mathematics
| Science
| Subjects
| Books
General
| Physics
| Science
| Subjects
| Books
Statistics
| Applied
| Mathematics
| Professional Science
| Professional & Technical
| Subjects
| Books
General
| Physics
| Professional Science
| Professional & Technical
| Subjects
| Books
General
| Education
| Professional & Technical
| Subjects
| Books
All Titles
| Qualifying Textbooks - Fall 2007
| Stores
| Books
ASIN: 0521774799 |
Book Description
The effort to build machines that are able to learn and undertake tasks such as datamining, image processing and pattern recognition has led to the development of artificial neural networks in which learning from examples may be described and understood. The contribution to this subject made over the past decade by researchers applying the techniques of statistical mechanics is the subject of this book. The authors provide a coherent account of various important concepts and techniques that are currently only found scattered in papers, supplement this with background material in mathematics and physics, and include many examples and exercises.
Download Description
Learning is one of the things that humans do naturally, and it has always been a challenge for us to understand the process. Nowadays this challenge has another dimension as we try to build machines that are able to learn and to undertake tasks such as datamining, image processing and pattern recognition. We can formulate a simple framework, artificial neural networks, in which learning from examples may be described and understood. The contribution to this subject made over the last decade by researchers applying the techniques of statistical mechanics is the subject of this book. The authors provide a coherent account of various important concepts and techniques that are currently only found scattered in papers, supplement this with background material in mathematics and physics and include many examples and exercises to make a book that can be used with courses, or for self-teaching, or as a handy reference.
Average customer rating:
|
Statistical Learning Theory and Stochastic Optimization (Lecture Notes Series)
O. Catoni
Manufacturer: Springer
ProductGroup: Book
Binding: Paperback
General
| Artificial Intelligence
| Computer Science
| Computers & Internet
| Subjects
| Books
Computer Mathematics
| Artificial Intelligence
| Computer Science
| Computers & Internet
| Subjects
| Books
General
| Science
| Subjects
| Books
General
| Mathematics
| Science
| Subjects
| Books
Number Systems
| Mathematics
| Science
| Subjects
| Books
Probability & Statistics
| Applied
| Mathematics
| Science
| Subjects
| Books
Statistics
| Applied
| Mathematics
| Professional Science
| Professional & Technical
| Subjects
| Books
Number Systems
| Mathematics
| Professional Science
| Professional & Technical
| Subjects
| Books
All Amazon Upgrade
| Amazon Upgrade
| Stores
| Books
Computers & Internet
| Amazon Upgrade
| Stores
| Books
Professional & Technical
| Amazon Upgrade
| Stores
| Books
Science
| Amazon Upgrade
| Stores
| Books
All Titles
| Qualifying Textbooks - Fall 2007
| Stores
| Books
Computers & Internet
| Qualifying Textbooks - Fall 2007
| Stores
| Books
Professional
| Qualifying Textbooks - Fall 2007
| Stores
| Books
Science
| Qualifying Textbooks - Fall 2007
| Stores
| Books
ASIN: 3540225722 |
Book Description
Statistical learning theory is aimed at analyzing complex data with necessarily approximate models. This book is intended for an audience with a graduate background in probability theory and statistics. It will be useful to any reader wondering why it may be a good idea, to use as is often done in practice a notoriously "wrong'' (i.e. over-simplified) model to predict, estimate or classify. This point of view takes its roots in three fields: information theory, statistical mechanics, and PAC-Bayesian theorems. Results on the large deviations of trajectories of Markov chains with rare transitions are also included. They are meant to provide a better understanding of stochastic optimization algorithms of common use in computing estimators. The author focuses on non-asymptotic bounds of the statistical risk, allowing one to choose adaptively between rich and structured families of models and corresponding estimators. Two mathematical objects pervade the book: entropy and Gibbs measures. The goal is to show how to turn them into versatile and efficient technical tools, that will stimulate further studies and results.
Book Description
Bayesian networks and decision graphs are formal graphical languages for representation and communication of decision scenarios requiring reasoning under uncertainty. Their strengths are two-sided. It is easy for humans to construct and to understand them, and when communicated to a computer, they can easily be compiled. Furthermore, handy algorithms are developed for analyses of the models and for providing responses to a wide range of requests such as belief updating, determining optimal strategies, conflict analyses of evidence, and most probable explanation. The book emphasizes both the human and the computer side. Part I gives a thorough introduction to Bayesian networks as well as decision trees and infulence diagrams, and through examples and exercises, the reader is instructed in building graphical models from domain knowledge. This part is self-contained and it does not require other background than standard secondary school mathematics. Part II is devoted to the presentation of algorithms and complexity issues. This part is also self-contained, but it requires that the reader is familiar with working with texts in the mathematical language. The author also: *Provides a well-founded practical introduction to Bayesian networks, decision trees and influence diagrams *Gives several examples and exercises exploiting the computer systems for Bayesian netowrks and influence diagrams *Gives practical advice on constructiong Bayesian networks and influence diagrams from domain knowledge. *Embeds decision making into the framework of Bayesian networks *Presents in detail the currently most efficient algorithms for probability updating in Bayesian networks *Discusses a wide range of analyes tools and model requests together with algorithms for calculation of responses. *Gives a detailed presentation of the currently most efficient algorithm for solving influence diagrams.
Finn V. Jensen is professor of computer science at the University of Aalborg.
Customer Reviews:
Good Book.......2006-03-01
For an introduction to the subject, this book is unequivocal in my experience with the literature. Great read that has propelled me forward into combining a bayesian network with a physical model to approach a very complex sediment transport problem.
A very good introduction to Bayesian networks.......2003-06-15
I am very pleased to have found a book that gives a modern, sound, and self-contained introduction to Bayesian networks. The only prerequisite is basic knowledge of probability. This makes sense because a Bayesian network is essentially a directed graph whose vertex set is a collection of random variables, while an edge from one variable X to another variable Y represents a belief that X has a causative effect on Y. For example, X could be the pregnancy status of a cow, while Y could be a blood test administered to the cow. Vertex Y would contain a contingency table that reflects the conditional probability of Y in terms of X. The author does well in explaining this, as well as adequately treating many of the practical issues surrounding Bayesian networks, such as design issues, network learing and tuning, and some basic algorithms (e.g. bucket elimination and junction trees) that aid in the efficient updating of variable probabilities due to new evidence that may instantiate or change the distribution of one or more variables.
The author also provides a good introduction to decision graphs, a close relative of Bayesian networks.
The aspect of Bayesian networks that I find most attractive is the fact that there is a "rational" way of designing a network, based on hypothesis, informational, and mediating variables, and their "causal" relationships. Unlike neural networks in which one is almost forced to guess the appropriate structure of the network, every node in a Bayesian network correpsonds with a state or quantity that can be measured either directly or indirectly through other variables. Thus, changes in a system model should only induce local changes in a Bayesian network, where as system changes might require the design and training of an entirely new neural network.
Another aspect of Bayesian networks that I find very compelling is the way in which they seem quite amendable to learning and the presentation of new evidence. This is true since knowledge updating is done locally (through variables), while the effects of those changes are witnessed globally through appropriate belief-updating algorithms.
On the downside, it should be noted that the operation of belief-updating is in general NP-hard, thus there exists a valid concern about the computational efficiency of Bayesian networks. Contrast this with the fact that once a nueral network has been trained, it is quite easy to compute. One would hope that these concerns will subside with more research, for the above mentioned benefits of Bayesian networks leads me to believe that these networks will have quite an influence on the future directions of machine learning.
Although this book will not go down in history as the definitive reference for Bayesian networks, it serves as a good conduit for explaining this quite interesting area of learning at a time when such few complete and modern references exist.
A lot about very little.......2003-05-06
The book covers many topics, but doesn't really cover them well. I would not recommend this book. I have learned litte from it.
Accessible introduction to Bayesian Networks.......2003-01-21
Among currently available introduction to Bayesian networks (also known as Bayes Net, Bayesian Belief Nets), this book is probably one of the most accessible. The book is divided into part I and II. Part I is intended for BN users (practitioners) and Part II more towards BN developers and researchers, as it contains algorithmic introduction of BN.
Prerequisites of the book as stated in the preface include Graph Theory and Calculus, both at introductory level. I personally did not have exposure to Graph theory, but I was able to understand most of the material without any help. Necessary probability theory is developed, but basic probability knowledge is also a prerequisite to digest the material to a reader without prior exposure of Probability as it shapes the core of the material in the book.
The strength of this text is in Part I where the author provides several examples to illustrate use of Bayesian Networks, Influence Diagrams and other models. I find it useful Influence Diagram as an extension of Bayesian Networks.
Most answers to Exercises at the end of each chapter are provided at the author's homepage, except answers of the last chapter. Answers that require graphical modeling software are also provided in Hugin format. (Hugin Lite can be downloaded from Hugin site.)
The downsides are that writing of the text is somewhat awkward, obscuring readers from understanding, that model building chapter could have been discussed more thoroughly, that material in Learning is barely present, and that definitions are sometimes not introduced upon the first encounter but they appear later in chapters. More different and complex examples could have been discussed to illustrate the material. Note: the author provides a page for Learning at his homepage.
Although this is an introduction to Bayesian Networks and Influence Diagrams, a reader should be equipped with some level of abstract thinking in order to digest the material.
This book is suitable for self-study. It has motivations for the uninitiated. References are provided at the end of the book and I was able to find some of them online. A notable is "A tutorial on Learning with Bayesian Networks" by Heckerman, to fill in the part of Learning in this book.
Other books at this level from users' perspective are:
Edwards, Introduction to Graphical Modeling (Utilizes software MIM.)
Clemen, et al., Making Hard Decisions (Uses Palisade Decision Tools suite. The book discusses Influence Diagrams but not Bayesian Networks.)
Further studies after completion of this book include:
Cowell, et al., Probabilistic Networks and Expert Systems
Lauritzen, Graphical Models
Pearl, Probabilistic Reasoning in Intelligent Systems
Pearl, Causality
Not worth the money.......2002-12-31
Chapter 1 is a nice introduction to probability. Chapter 2 is readable. Chapter 3 is poorly presented, and you feel sad for having wasted so much money on a book with only one intelligible chapter.
Books:
- Statistical Tables for the Design of Clinical Studies
- Stranger Than Fiction: True Stories
- The Antipodes of the Mind: Charting the Phenomenology of the Ayahuasca Experience
- The Bat House Builder's Handbook, Completely Revised and Updated
- The Brain That Changes Itself: Stories of Personal Triumph from the Frontiers of Brain Science (James H. Silberman Books)
- The Coming Plague: Newly Emerging Diseases in a World Out of Balance
- The Developing Person Through the Life Span (paper)
- The Devil's Teeth: A True Story of Obsession and Survival Among America's Great White Sharks
- The Emotion Machine: Commonsense Thinking, Artificial Intelligence, and the Future of the Human Mind
- The Enormous Egg
Books Index
Books Home
Recommended Books
- Personal History
- Real Parents Real Children
- I Am Madame X: A Novel
- Gunfitting: The Quest for Perfection for Shotguns And Rifles
- History: Fiction or Science
- History: Fiction or Science
- I Am a Strange Loop
- Unsportsmanlike Conduct: Exploiting College Athletes
- Hayek's Journey: The Mind of Friedrich Hayek
- The Military Memoirs of General John Pope
The Nature of Statistical Learning Theory (STATISTICS FOR ENGINEERING AND INFORMATION SCIENCE)
The Elements of Statistical Learning
An Introduction To Support Vector Machines: And Other Kernel-Based Learning Methods
Learning with Kernels: Support Vector Machines, Regularization, Optimization, and Beyond (Adaptive Computation and Machine Learning)
Pattern Recognition and Machine Learning (Information Science and Statistics)
