Clustering of moving vectors for evolutionary computation

Abstract

We propose a method for clustering moving vectors oriented around two different local optima and some methods for improving the clustering performance. Evolutionary computation is an optimization method for finding the global optimum iteratively using multiple individuals; we propose a method for estimating the global optimum mathematically using the moving vectors between parent individuals and their offspring. Our proposed clustering method is the first to tackle the extension of the estimation method to multi-modal optimization. We describe the algorithm of the clustering method, the improvements made to the method, and the estimation performance for two local optima.