» » Probabilistic Graphical Models: Principles and Techniques (Adaptive Computation and Machine Learning)

Download Probabilistic Graphical Models: Principles and Techniques (Adaptive Computation and Machine Learning) fb2

by Francis Bach,Daphne Koller
Download Probabilistic Graphical Models: Principles and Techniques (Adaptive Computation and Machine Learning) fb2
Computer Science
  • Author:
    Francis Bach,Daphne Koller
  • ISBN:
  • ISBN13:
  • Genre:
  • Publisher:
    The MIT Press; 1 edition (July 31, 2009)
  • Pages:
    1270 pages
  • Subcategory:
    Computer Science
  • Language:
  • FB2 format
    1771 kb
  • ePUB format
    1361 kb
  • DJVU format
    1751 kb
  • Rating:
  • Votes:
  • Formats:
    mobi lit mobi lrf

A general framework for constructing and using probabilistic models of complex systems that would enable a computer to use available information for making decisions.

Most tasks require a person or an automated system to reason―to reach conclusions based on available information. The framework of probabilistic graphical models, presented in this book, provides a general approach for this task. The approach is model-based, allowing interpretable models to be constructed and then manipulated by reasoning algorithms. These models can also be learned automatically from data, allowing the approach to be used in cases where manually constructing a model is difficult or even impossible. Because uncertainty is an inescapable aspect of most real-world applications, the book focuses on probabilistic models, which make the uncertainty explicit and provide models that are more faithful to reality.

Probabilistic Graphical Models discusses a variety of models, spanning Bayesian networks, undirected Markov networks, discrete and continuous models, and extensions to deal with dynamical systems and relational data. For each class of models, the text describes the three fundamental cornerstones: representation, inference, and learning, presenting both basic concepts and advanced techniques. Finally, the book considers the use of the proposed framework for causal reasoning and decision making under uncertainty. The main text in each chapter provides the detailed technical development of the key ideas. Most chapters also include boxes with additional material: skill boxes, which describe techniques; case study boxes, which discuss empirical cases related to the approach described in the text, including applications in computer vision, robotics, natural language understanding, and computational biology; and concept boxes, which present significant concepts drawn from the material in the chapter. Instructors (and readers) can group chapters in various combinations, from core topics to more technically advanced material, to suit their particular needs.

This popular book makes a noble attempt at unifying the many different types of probabilistic models used in artificial intelligence. It seems like a good reference manual for people who are already familiar with the fundamental concepts of commonly used probabilistic graphical models. However, it contains a lot of rambling and jumping between concepts that will quickly confuse a reader who is not already familiar with the subject. While the book appears to be systematic in introducing the subject with mathematical rigor (definitions and theorems), it actually skips a lot of fundamental concepts and leaves a lot of important proofs as exercises. I would recommend that a beginner in the subject start with another book like that by Jordan and Bishop, while keeping this book around as a reference manual or bank of practice problems for further study. The Coursera class on this subject is much easier to follow than this book is.
This is a great book on the topic, regardless of whether you are new to probabilistic graphical models or have some familiarity with them but would like a deeper exploration of theory and/or implementation. I have read a number of books and papers on this topic (including Barber's and Bishop's) and I much prefer this one. Dr. Koller's style of writing is to start with simple theory and examples and walk the reader up to the full theory, while adding reminders of relevant topics covered elsewhere. She accomplishes this without condescending to or belittling the reader, or being overly verbose; each of the 1200 pages is concise and well edited. There is an OpenClassroom course that accompanies the book (CS 228), which I highly recommend viewing, as it contains that same style of teaching but in a different format and often with a somewhat different approach.
companion text for my coursera class
Lahorns Gods
Judging by the first few chapters, the text is cumbersome and not as clear as it could have been under a more disciplined writing style; Sentences and paragraphs are longer than they should be, and the English grammar is most of the time improper or just a little odd. Reads too much like a transcript of a free speech lecture. Hopefully this alleviates later on in the book.
Very usefull book, and te best. conpanion for the course about.
I bought this book to use for the Coursera course on PGM taught by the author. It was essential to being able to follow the course. I would not say that it is an easy book to pick up and learn from. It was a good reference to use to get more details on the topics covered in the lectures.
with an eye to taking the course. Very informative. Although the phrase "in context" covers a multitude of sins. I'd prefer the distinction between the the distribution of an intersection of random variables (where comma's are used as a short-hand) and joint distributions a bit clearer.

Aside, I managed to find an error not listed on the errata web page for the book. The equation for MAP queries on page 26 has it as the maximal assignment of a JOINT distribution, while on the next page it is the maximal assignment of a CONDITIONAL distribution (I believe this is the correct one). This was a little confusing until I read page 26 a bit closer.

Before you ask, yes I do read Math textbooks for pleasure.
I just started reading this book for a course I want to do the coming semester. It seems like a very interesting book, exploring in depth probabilistic graphical models. Some topics are pretty tough to understand, and require additional background knowledge (such as statistics, algorithms, machine learning and AI). Another issue is that the book doesn't come with answers to the exercises, so you never know if you are on the right track when solving them.