On the anytime behavior of IPOP-CMA-ES

Authors:
Manuel López-Ibáñez;Tianjun Liao;Thomas Stützle
Affiliations:
IRIDIA, CoDE, Université Libre de Bruxelles, Brussels, Belgium;IRIDIA, CoDE, Université Libre de Bruxelles, Brussels, Belgium;IRIDIA, CoDE, Université Libre de Bruxelles, Brussels, Belgium
Venue:
PPSN'12 Proceedings of the 12th international conference on Parallel Problem Solving from Nature - Volume Part I
Year:
2012

Citing 6
Cited 1

Stochastic Local Search: Foundations & Applications

Stochastic Local Search: Foundations & Applications
Completely Derandomized Self-Adaptation in Evolution Strategies

Evolutionary Computation
An experimental investigation of model-based parameter optimisation: SPO and beyond

Proceedings of the 11th Annual conference on Genetic and evolutionary computation
Improvement strategies for the F-Race algorithm: sampling design and iterative refinement

HM'07 Proceedings of the 4th international conference on Hybrid metaheuristics
Tuning parameters across mixed dimensional instances: a performance scalability study of Sep-G-CMA-ES

Proceedings of the 13th annual conference companion on Genetic and evolutionary computation
Performance assessment of multiobjective optimizers: an analysis and review

IEEE Transactions on Evolutionary Computation

Testing the impact of parameter tuning on a variant of IPOP-CMA-ES with a bounded maximum population size on the noiseless BBOB testbed

Proceedings of the 15th annual conference companion on Genetic and evolutionary computation

Quantified Score

Hi-index	0.00

Visualization

Abstract

Anytime algorithms aim to produce a high-quality solution for any termination criterion. A recent proposal is to improve automatically the anytime behavior of single-objective optimization algorithms by incorporating the hypervolume, a well-known quality measure in multi-objective optimization, into an automatic configuration tool. In this paper, we show that the anytime behavior of IPOP-CMA-ES can be significantly improved with respect to its default parameters by applying this method. We also show that tuning IPOP-CMA-ES with respect to the final quality obtained after a large termination criterion leads to better results at that particular termination criterion, but worsens the performance of IPOP-CMA-ES when stopped earlier. The main conclusion is that IPOP-CMA-ES should be tuned with respect to the anytime behavior if the exact termination criterion is not known in advance.