Benchmarking a BI-Population CMA-ES on the BBOB-2009 Noisy Testbed

Nikolaus Hansen (1, 2)

1 TAO - Machine Learning and Optimisation
2 MSR - INRIA - Microsoft Research - Inria Joint Centre

We benchmark the BI-population CMA-ES on the BBOB-2009 noisy functions testbed. BI-population refers to a multistart strategy with equal budgets for two interlaced restart strategies, one with an increasing population size and one with varying small population sizes. The latter is presumably of little use on a noisy testbed. The BI-population CMA-ES could solve 29, 27 and 26 out of 30 functions in search space dimension 5, 10 and 20 respectively. The time to find the solution ranges between $100 D$ and $10^5 D^2$ objective function evaluations, where $D$ is the search space dimension.

Domains: Machine Learning [cs.LG]

Nikolaus Hansen. Benchmarking a BI-Population CMA-ES on the BBOB-2009 Noisy Testbed. ACM-GECCO Genetic and Evolutionary Computation Conference, Jul 2009, Montreal, Canada. ⟨inria-00382101⟩