Thursday, July 27, 2006

 

EDAs and cross-entropy methods

Estimation of distribution algorithms (EDAs)---also known as probabilistic model-building genetic algorithms (PMBGAs) and iterated density estimation algorithms (IDEAs)---are population-based optimization methods that use probabilistic modeling to guide the search for the optimum. In other words, we may say that EDAs replace standard variation operators of genetic and evolutionary algorithms by building and sampling a probabilistic model of promising solutions found so far. EDAs date back to 1994 when Shumeet Baluja proposed PBIL algorithm, and as most of us already know, since then the field has made a significant progress.

A similar approach has been taken in 1997 by Reuven Rubinstein, who proposed the cross-entropy method to provide an efficient tool for rare-event simulation and optimization. Just like EDAs, CE proceeds by generating samples of points from a probabilistic model and then updating the probabilistic model to better fit high-quality points in the generated sample. For me the most important question is: what can we learn from each other?

Comments: Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?