Illinois Genetic Algorithms Laboratory

Tsutsui, S., Goldberg, D.E., Sastry, K. (2000)
TR No.: 2000033 | Download PDF | Download PS

Abstract:
In recent years, many researchers have concentrated on using real-valued genes in genetic and evolutionary algorithms (GEAs). Previous studies have proposed simplex crossover (SPX) for real-coded GAs. SPX has several good characteristics and works well on various test functions. However, SPX fails on functions that consist of tightly linked sub-functions. On those functions, SPX should be applied on each tightly linked group of parameters. In this paper, we propose
a method of linkage identification in real-coded GAs with SPX and evaluate it using several test functions. The mechanism works well on many of the test functions used. We also discuss difficulties with the proposed method on more complex
test functions and show possible solutions to the problems.