Identifying variable interaction using mutual information of multiple local optima
Yapei Wu, Xingguang Peng, Demin Xu · Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2019
Identifying the interaction of search variables of black-box optimization problem is beneficial for optimization task. However, very little research pay attention to the quality of information source, i.e. what information is beneficial for identifying the interactions between variables. In this paper, we propose a new method that utilizes multiple local optima as information sources to identify the interaction between variables. First, a multimodal optimization algorithm is used to search for multiple local optima of the optimization problem. Then, hierarchical clustering is used to cluster and discretize local optima. Finally, the interaction between variables is quantified using the mutual information of local optima. Experimental results on three 12-dimensional multimodal problems show that the proposed method can effectively identify the interactions among decision variables.