Add to Chrome
Add to Firefox
Add to EdgeAdaptive strategy in differential evolution via explicit exploitation and exploration controls
Feb 03, 2020
When introducing new strategies to the existing one, two key issues should be addressed. One is to efficiently distribute computational resources so that the appropriate strategy dominates. The other is to remedy or even eliminate the drawback of inappropriate strategies. Adaptation is a popular and efficient method for strategy adjustments and has been widely studied in the literature. Existing methods commonly involve the trials of multiple strategies and then reward better-performing one with more resources based on their previous performance. As a result, it may not efficiently address those two key issues. On the one hand, they are based on trial-and-error with inappropriate strategies consuming resources. On the other hand, since multiple strategies are involved in the trial, the inappropriate strategies could mislead the search. In this paper, we propose an adaptive differential evolution (DE) with explicit exploitation and exploration controls (Explicit adaptation DE, EaDE), which is the first attempt using offline knowledge to separate multiple strategies to exempt the optimization from trial-and-error. EaDE divides the evolution process into several SCSS (Selective-candidate with similarity selection) generations and adaptive generations. Exploitation and exploration needs are learned in the SCSS generations by a relatively balanced strategy. While in the adaptive generations, to meet these needs, two other alternative strategies, an exploitative one or an explorative one is employed. Experimental studies on 28 benchmark functions confirm the effectiveness of the proposed method.
