THEORETICAL ANALYSIS OF THE GA PERFORMANCE WITH A MULTIPLICATIVE ROYAL ROAD fUNCTION

摘要

The performance of genetic algorithms (GAs) is theoretically estimated with multiplicative royal-road functions (mRR-functions). Using a macro-schema analysis, the effects of selection, mutation, and crossover are quantitatively estimated, which enables formulation of the innovation time and takeover time of component schemata as a function of genetic parameters. Theoretical estimation is compared to the experimental results of a simple GA, and it is shown that the theoretical results are in good agreement with experimental ones specifically when the innovation time is much larger than the takeover time.