Estimation of Distribution Algorithms: A New Tool for Evolutionary Computation

Rating:
Summary: Many topics covered, some chapters are a little weak
Review: This is a good book to learn about Estimation of Distribution Algorithms (EDAs or also called DEAs or Iterated DEAs). These algorithms are similar to evolutionary algorithms, but do not use the crossover or mutation operators of evolutionary search. EDAs instead create a probabilistic model of good solutions and use the model to generate new search points. It's a nifty idea and it works.

Most of the chapters of this edited collection were authored or coauthored by the editors. So, algorithms developed by other people do not get a lot of attention. However, the editors (or is it the authors) manage to include chapters on combinatorial, continuous, and discrete optimization.

There is a section on machine learning applications that is OK, but the last chapter on training neural nets with EDAs is very weak (look ma I used this and it worked...). Except for this chapter, the rest of the chapters in this section use careful experiments and statistics to make their points.

Making the source code available would have improved things and would make it easier for people to try these algorithms.