Towards the Role of Heuristic Knowledge in EA

摘要

Evolutionary Algorithm (EA) is a stochastic search algorithm and widely used in various real world problems. Classic EA uses little problem specific knowledge, so it is called lean knowledge approach. Because of the randomicity of crossover, mutation and selection, its' searching strategy is semi-blind, and the efficiency is usually low. In order to acquire an efficient and effective EA that suits difficult real-world problems, we try to best incorporate heuristic knowledge into an EA to guide the search focusing on the most promising area. By comparing different Eas for solving the traveling sales man problem (TSP) and auto-generating test paper problem, we investigate the role of heuristic knowledge in EA.