A more efficient rank-one covariance matrix update for evolution strategies

11 Citationer (Scopus)

Abstract

Learning covariance matrices of Gaussian distributions is at the heart of most variable-metric randomized algorithms for continuous optimization. If the search space dimensionality is high, updating the covariance or its factorization is computationally expensive. Therefore, we adopt an algorithm from numerical mathematics for rank-one updates of Cholesky factors. Our methods results in a quadratic time covariance matrix update scheme with minimal memory requirements. The numerically stable algorithm leads to triangular Cholesky factors. Systems of linear equations where the linear transformation is defined by a triangular matrix can be solved in quadratic time. This can be exploited to avoid the additional iterative update of the inverse Cholesky factor required in some covariance matrix adaptation algorithms proposed in the literature. When used together with the (1+1)-CMA-ES and the multi-objective CMA-ES, the new method leads to a memory reduction by a factor of almost four and a faster covariance matrix update. The numerical stability and runtime improvements are demonstrated on a set of benchmark functions.

OriginalsprogEngelsk
TitelProceedings of the 2015 ACM Conference on Foundations of Genetic Algorithms XIII
Antal sider8
ForlagAssociation for Computing Machinery
Publikationsdato17 jan. 2015
Sider129-136
ISBN (Trykt)978-1-4503-3434-1
DOI
StatusUdgivet - 17 jan. 2015
BegivenhedACM Conference on Foundations of Genetic Algorithms 2015 - Aberystwyth, Wales, Storbritannien
Varighed: 17 jan. 201520 jan. 2015
Konferencens nummer: 13

Konference

KonferenceACM Conference on Foundations of Genetic Algorithms 2015
Nummer13
Land/OmrådeStorbritannien
ByAberystwyth, Wales
Periode17/01/201520/01/2015

Emneord

  • cholesky factorization, cma-es, covariance matrix adaptation, rank-one update

Fingeraftryk

Dyk ned i forskningsemnerne om 'A more efficient rank-one covariance matrix update for evolution strategies'. Sammen danner de et unikt fingeraftryk.

Citationsformater