Adapting the Covariance Matrix in Evolution Strategies
Silja Meyer-Nieberg, Erik Kropat · 2014
Evolution strategies belong to the best performing modern metaheuristics for continuous optimization. This paper addresses the covariance matrix adaptation in evolution strategies which is central to the algorithm. Nearly all approaches so far consider the sample covariance matrix as one of the main factors for the adaptation. However as known from modern statistics, this estimate may be of poor quality in many cases. Unfortunately, these cases are encountered often in practical applications of evolution strategies. This paper explores the use of different estimates, which have not been used before.