Advertisement

Applications of CE to Machine Learning

  • Reuven Y. Rubinstein
  • Dirk P. Kroese
Part of the Information Science and Statistics book series (ISS)

Abstract

In this chapter we apply the CE method to several problems arising in machine learning, specifically with respect to optimization. In Section 8.1, adapted from [50], we apply CE to the well-known mastermind game. Section 8.2, based partly on [112], describes the application of the CE method to Markov decision processes. Finally, in Section 8.3 the CE method is applied to clustering problems. In addition to its simplicity, the advantage of using the CE method for machine learning is that it does not require direct estimation of the gradients, as many other algorithms do (for example, the stochastic approximation, steepest ascent, or conjugate gradient method). Moreover, as a global optimization procedure the CE method is quite robust with respect to starting conditions and sampling errors, in contrast to some other heuristics, such as simulated annealing or guided local search.

Keywords

Markov Decision Process Vector Quantization Stochastic Approximation Policy Search Mastermind Game 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

Copyright information

© Springer Science+Business Media New York 2004

Authors and Affiliations

  • Reuven Y. Rubinstein
    • 1
  • Dirk P. Kroese
    • 2
  1. 1.Department of Industrial Engineering and ManagementTechnionTechnion City, HaifaIsrael
  2. 2.Department of MathematicsUniversity of QueenslandBrisbaneAustralia

Personalised recommendations