Stochastic Approximation

ISBN-13: 9780521515924 / Angielski / Twarda / 2008 / 176 str.

Vivek S. Borkar

Opis

This simple, compact toolkit for designing and analyzing stochastic approximation algorithms requires only a basic understanding of probability and differential equations. Although powerful, these algorithms have applications in control and communications engineering, artificial intelligence and economic modeling. Unique topics include finite-time behavior, multiple timescales and asynchronous implementation. There is a useful plethora of applications, each with concrete examples from engineering and economics. Notably it covers variants of stochastic gradient-based optimization schemes, fixed-point solvers, which are commonplace in learning algorithms for approximate dynamic programming, and some models of collective behavior.

Recenzje

'This simple compact toolkit for designing and analyzing stochastic approximation algorithms requires only basic literacy in probability and differential equations … Ideal for graduate students, researchers and practitioners in electrical engineering and computer science, especially those working in control, communications, signal processing and machine learning, this book is also relevant to economics, probability and statistics.' L'Enseignement Mathématique

Spis treści

Preface; 1. Introduction; 2. Basic convergence analysis; 3. Stability criteria; 4. Lock-in probability; 5. Stochastic recursive inclusions; 6. Multiple timescales; 7. Asynchronous schemes; 8. A limit theorem for fluctuations; 9. Constant stepsize algorithms; 10. Applications; 11. Appendices; References; Index.