A New Framework for Self-adapting Control Parameters in Multi-objective Optimization

摘要: Proper tuning of control parameters is critical to the performance a multi-objective evolutionary algorithm (MOEA). However, developments methods for optimization are insufficient compared single-objective optimization. To circumvent this issue, paper proposes novel framework that can self-adapt parameter values from an objective-based perspective. Optimal parametric setups each objective will be efficiently estimated by combining with grouping mechanism. Subsequently, position information individuals in space utilized achieve more efficient adaptation among multiple objectives. The new implemented into two classical Differential-Evolution-based MOEAs help adapt scaling factor F objective-wise manner. Three state-of-the-art applied respectively validate robustness proposed mechanisms. Experimental results demonstrate effective and robust solving problems.