Genetic algorithms: combining evolutionary and 'non'-evolutionary methods in tracking dynamic global optima
Simon M. Garrett, Joanne H. Walker · 2002
The ability to track dynamic functional optima is important in many practical tasks. Recent research in this area has concentrated on modifying evolutionary algorithms (EAs) by triggering changes in control parameters, ensuring population diversity, or remembering past solutions. A set of results are presented that favourably compare hill climbing with a genetic algorithm, and reasons for the results are suggested. A method is then introduced, Evolutionary Random Search (ERS), that combines crossover and hill climbing mutation in a novel manner. It is assessed against the GA and hill climbing tests, and the encouraging results are discussed.