A Two Phase Approach Based on Dynamic Variable Grouping and Self-Adaptive Group Search for Large Scale Optimization

摘要

In this paper, a self-adaptive two phase approach for large scale optimization is proposed. In the first phase, we design a uniform discrete search method which can quickly and roughly scan the search space and find good initial points. Then we continuously narrow the search space and make more precise search in a dynamically self-adaptive way. In the second phase, we design a dynamically self-adaptive grouping search scheme which can group the variables into several groups dynamically and assign different function evaluations to different variable groups self-adaptively during each group search. The experiment results indicate the proposed algorithm is effective and efficient.