Crossing the road to efficient IDEAs for permutation problems

Abstract

In this paper, we introduce the ICE framework in which crossover from genetic algorithms (GAs) is incorporated in iterated density estimation evolutionary algorithms (IDEAs). We focus on permutation optimization problems and show how pure continuous IDEAs can be applied using the random keys representation. The problems that are hereby encountered, motivate the use of ICE As a result, permutation linkage information is effectively processed, resulting in efficient optimization of deceptive permutation problems of a bounded order. Experiments show that ICE outperforms pure continuous IDEAs. Furthermore, we show that ICE gives insight into how new IDEAs can be designed that efficiently work directly in the permutation search space.