LNCS Homepage
CD ContentsAuthor IndexSearch

Ruffled by Ridges: How Evolutionary Algorithms Can Fail

Darrell Whitley, Monte Lunacek, and James Knight

Computer Science, Colorado State University, Fort Collins, CO 80523

Abstract. The representations currently used by local search and some evolutionary algorithms have the disadvantage that these algorithms are partially blind to “ridges” in the search space. Both heuristics search and gradient search algorithms can exhibit extremely slow convergence on functions that display ridge structures. A class of rotated representations are proposed and explored; these rotated representations can be based on Principal Components Analysis, or use the Gram-Schmidt orthogonalization method. Some algorithms, such as CMA-ES, already make use of similar rotated representations.

LNCS 3103, p. 294 ff.

Full article in PDF


lncs@springer.de
© Springer-Verlag Berlin Heidelberg 2004