Population Control meets Doob's Martingale Theorems: the Noise-free Multimodal Case

TL;DR

This paper introduces a population size adaptation method for multimodal optimization that effectively escapes local minima without restarts, supported by theoretical proofs and experimental validation.

Contribution

It demonstrates that TBPSA, combined with a naive recommendation strategy, can reliably escape plateaus in noise-free multimodal optimization, backed by theoretical analysis.

Findings

TBPSA can escape plateaus with probability one.

The naive recommendation strategy is effective in multimodal settings.

The method avoids the need for random restarts.

Abstract

We study a test-based population size adaptation (TBPSA) method, inspired from population control, in the noise-free multimodal case. In the noisy setting, TBPSA usually recommends, at the end of the run, the center of the Gaussian as an approximation of the optimum. We show that combined with a more naive recommendation, namely recommending the visited point which had the best fitness value so far, TBPSA is also powerful in the noise-free multimodal context. We demonstrate this experimentally and explore this mechanism theoretically: we prove that TBPSA is able to escape plateaus with probability one in spite of the fact that it can converge to local minima. This leads to an algorithm effective in the multimodal setting without resorting to a random restart from scratch.