A Comparison of Evolution Strategies with Other Direct Search Methods in the Presence of Noise

摘要: Evolution strategies are general, nature-inspired heuristics for search and optimization. Due to their use of populations candidate solutions advanced adaptation schemes, there is a common belief that evolution especially useful optimization in the presence noise. Empirical evidence as well number theoretical findings with respect performance on class spherical objective functions disturbed by Gaussian noise support belief. However, little known capabilities relative those other direct strategies. In present paper, results summarized discussed. Then, compared empirically several noisy, environment have been obtained in. simplicity environment, easily interpretable can serve reveal respective strengths weaknesses algorithms. It seen low levels noise, most exhibit similar degrees efficiency. For higher step length scheme affords greater degree robustness than algorithms tested.