Weight Adaptation for Improving Parallel Performance of Adaptive Stochastic Natural Gradient

2026-06-18

AI Summary

This work addresses the ambiguity in weight parameter configuration within Adaptive Stochastic Natural Gradient (ASNG) under large-population parallel evaluation and proposes a Weight-Adaptive ASNG (WA-ASNG) method. WA-ASNG is the first to decouple weights from the learning rate and adaptively optimize them separately: weights are dynamically adjusted via gradient ascent to maximize the expected improvement of the objective function, while the learning rate ensures monotonic convergence. By integrating probabilistic model evolution, natural gradient estimation, and a dedicated weight-adaptation mechanism, WA-ASNG significantly outperforms PBIL and standard ASNG across various binary optimization problems—particularly with population sizes ranging from 25 to 100—and demonstrates robustness and efficiency even under strong noise conditions.

Abstract

Probabilistic model-based evolutionary algorithms are promising for black-box optimization. Specifically, the adaptive stochastic natural gradient (ASNG) adaptively updates its learning rate, a typical hyperparameter in probabilistic model-based evolutionary algorithms, thereby realizing efficient and robust optimization. Although weight parameters are common hyperparameters, with the increasing demand for parallel evaluation of time-consuming tasks, it remains unclear how to set suitable weights for larger population sizes. In this paper, we propose Weight Adaptation ASNG (WA-ASNG), which incorporates a weight adaptation mechanism into ASNG. We calculated the estimated signal of the update direction from the accumulations of the natural gradient. Then, to maximize the signal, WA-ASNG adaptively updates its weight parameters by a gradient ascent over the optimization. While the learning rate adaptation plays a role in satisfying a sufficient condition for monotonic improvement of the expected objective value, the mechanism of weight adaptation is intended to maximize this improvement. The experimental results demonstrate that WA-ASNG outperforms PBIL and ASNG across various settings with population sizes ranging from 25 to 100 for binary optimization problems. Furthermore, WA-ASNG can perform efficiently in the presence of strong noise. Our code is available at https://github.com/shiralab/WA-ASNG .

Problem

• weight adaptation
• population size
• parallel evaluation
• probabilistic model-based evolutionary algorithms
• black-box optimization

Innovation

• weight adaptation
• adaptive stochastic natural gradient
• probabilistic model-based evolutionary algorithm
• parallel black-box optimization
• natural gradient