Graphical Models for Machine Learning and Digital Communication
A variety of problems in machine learning and digital communication deal with complex but structured natural or artificial systems. In this book, Brendan Frey uses graphical models as an overarching framework to describe and solve problems of pattern classification, unsupervised learning, data compression, and channel coding. Using probabilistic structures such as Bayesian belief networks and Markov random fields, he is able to describe the relationships between random variables in these systems and to apply graph-based inference techniques to develop new algorithms. Among the algorithms described are the wake-sleep algorithm for unsupervised learning, the iterative turbodecoding algorithm (currently the best error-correcting decoding algorithm), the bits-back coding method, the Markov chain Monte Carlo technique, and variational inference.
HardcoverOut of Print ISBN: 9780262062022 211 pp. | 9 in x 6.2 in
This well-written monograph provides a stimulating and insightful perspective on the latest and greatest advances in capacity-approaching codes and other areas within the general framework of graphical models. Anyone doing research in these areas should read this book.
G. David Forney, Jr.
Vice President, Motorola
this is a wonderfully timely book and it will bring the reader up-to-date in the bewilderingly diverse yet excitingly unified world of algorithms on graphical models. It culminates, appropriately, with the best exposition of the astonishing turbo-codes of Berrou, Glaviuex, and Thitimajshima that has yet appeared.
Robert J. McEliece
Allen E. Puckett Professor of Electrical engineering, California Institute of Technology
This text provides a wonderfully elegant synthesis of ideas in learning algorithms and communication models, using probability theory (and more specifically, graphical models) as a common language. The clarity of presentation is excellent, clearly demonstrating how probabilistic modeling is a guiding principle underlying a very broad canvas of 'data interpretation' algorithms. The unifying theme of graphical models allows Frey to present many different ideas in a very effective and succinct manner, making the text a very valuable read for both newcomers and seasoned researchers.
Information and Computer Science, University of California at Irvine
The scope of this book is immense. Brendan Frey presents with great clarity the cutting edge of research on the learning of graphical models, the compression of data using latent variable models, and the channel coding. It will be appreciated by everyone from computer scientists and psychologists to statisticians, engineers, and information theorists.
Dr. David J.C. MacKay
department of Physics, University of Cambridge