A New Hybrid Algorithm for Solving Large Scale Global Optimization Problems

Abstract

For large scale global optimization (LSGO) problems, many algorithms have been proposed in recent years. However, there are still some issues to be further handled since the search space grows exponentially and the problem solving becomes more and more difficult as the problem scale becomes larger and larger. In this paper, we propose a new hybrid algorithm for solving large-scale global optimization problems. First, we adopt an existing group algorithm to divide the large-scale problem into several small-scale problems. Second, a modified self-adaptive discrete scan method is designed to roughly scan the whole search space and then focus the search on the promising regions. Third, a hybrid search strategy is proposed, which adaptively chooses the one-dimensional search scheme or the covariance matrix adaptation evolutionary strategy to solve the subproblems of separable, partially (additively) separable problems or non-separable problems, respectively. To demonstrate the performance of the proposed algorithm, we conduct the experiments on 15 difficult LSGO problems in CEC'2013 benchmark suite and compare the performance of the proposed algorithm with that of the several state-of-the-art algorithms. The results show that the proposed algorithm is more effective than the compared algorithms in terms of solution accuracy.