Generative Evolutionary Strategy For Black-Box Optimizations

TL;DR

This paper introduces a hybrid black-box optimization method combining evolution strategies and generative neural networks, effectively handling high-dimensional, multi-objective, and stochastic functions, outperforming existing methods.

Contribution

It presents a novel cooperative framework integrating ES and GNN for improved high-dimensional black-box optimization.

Findings

Outperforms baseline methods like ES and Bayesian optimization

Effective in high-dimensional, multi-objective, stochastic problems

Reliable training of surrogate networks achieved

Abstract

Many scientific and technological problems are related to optimization. Among them, black-box optimization in high-dimensional space is particularly challenging. Recent neural network-based black-box optimization studies have shown noteworthy achievements. However, their capability in high-dimensional search space is still limited. This study proposes a black-box optimization method based on the evolution strategy (ES) and the generative neural network (GNN) model. We designed the algorithm so that the ES and the GNN model work cooperatively. This hybrid model enables reliable training of surrogate networks; it optimizes multi-objective, high-dimensional, and stochastic black-box functions. Our method outperforms baseline optimization methods in this experiment, including ES, and Bayesian optimization.