Baldwin effect based self-adaptive generalized genetic algorithm
Youfa Sun, Feiqi Deng · 2005
Standard genetic algorithm conducts probabilistic parallel searches for the best chromosome by repeating generate-and-test processes, which completely ignore experiences gained during individuals' lifetime. Such inborn defect, however, is fully intact under conventional improvements. In this paper, a novel self-adaptive generalized GA based on Baldwin effect is proposed. A fourth operator Baldwin learning, is introduced. All members must perform Baldwin learning before they enter into the gene pool for further crossover and mutation. Besides, mechanisms of "inbreeding is forbidden" and activation are built in to solve problems of crowding and slow convergence. Finally, the kernel of inconsistent self-adaptive GA is also fused into this new algorithm. Application to a benchmark problem shows the new algorithm is feasible and highly effective.