The cross-entropy method. A unified approach to combinatorial optimization, Monte-Carlo simulation and machine learning. (Q4828558)

This book is a good introduction to the cross-entropy (CE) method, an approach to combinatorial optimization and rare-event simulation based on minimizing the cross-entropy (also known as the Kullback-Leibler distance) between a sampling distribution and an unknown target distribution. It is intended to be ``accessible to advanced undergraduate students and engineers who simply want to apply the CE method in their work, while at the same time accentuating the unifying and novel ideas behind the CE method.'' Uses of the CE method are illustrated with many simple examples, longer case studies, and an appendix of Matlab programs. Chapter 1 presents the theory behind the CE method, though many readers will elect to begin with Chapter 2, a tutorial introduction demonstrating the CE method's application to rare-event simulation and optimization. The remaining chapters address more thoroughly the application of the CE method to various problem classes, in simulation and machine learning, as well as combinatorial, continuous and noisy optimization. The authors also provide helpful ``rules-of-thumb'' for parameter tuning, and discuss self-tuning variants of the CE method.