A Simple Yet Efficient Rank One Update for Covariance Matrix Adaptation

TL;DR

This paper introduces a computationally efficient rank one update method for CMA-ES that simplifies the process by avoiding matrix decompositions, maintaining competitive performance.

Contribution

The paper presents a novel approximated rank one update for CMA-ES that reduces computational complexity while preserving effectiveness.

Findings

Outperforms Cholesky CMA-ES in experiments

Maintains competitive performance with simpler updates

Avoids costly matrix decompositions

Abstract

In this paper, we propose an efficient approximated rank one update for covariance matrix adaptation evolution strategy (CMA-ES). It makes use of two evolution paths as simple as that of CMA-ES, while avoiding the computational matrix decomposition. We analyze the algorithms' properties and behaviors. We experimentally study the proposed algorithm's performances. It generally outperforms or performs competitively to the Cholesky CMA-ES.