Design Of An Efficient Hyper-heuristic Algorithm Cma-vns For Combinatorial Black-box Optimization Problems

Abstract

We present a hyper-heuristic algorithm for solving combinatorial black-box optimization problems. The algorithm named CMA-VNS stands for a hybrid of variants of Covariance Matrix Adaptation Evolution Strategy (CMA-ES) and Variable Neighborhood Search (VNS). The framework design and the design profiles of variants of CMA-VNS are introduced to enhance the intensification of searching for conventional CMA-ES solvers. We explain the parameter configuration details, the heuristic profile selection, and the rationale of incorporating machine learning methods during the study. Experimental tests and the results of the first and the second Combinatorial Black-Box Optimization Competitions (CB-BOC 2015, 2016) confirmed that CMA-VNS is a competitive hyper-heuristic algorithm.