Black-Box Optimization with Local Generative Surrogates

TL;DR

This paper introduces a method for optimizing non-differentiable, stochastic black-box simulators by using local deep generative surrogate models to approximate gradients, enabling efficient gradient-based optimization.

Contribution

It presents a novel approach that employs local generative models to approximate simulator gradients, improving optimization speed over existing methods in certain cases.

Findings

Faster convergence to minima compared to Bayesian optimization and other methods.

Effective approximation of simulator gradients using local deep generative models.

Applicable to simulators with low-dimensional parameter dependencies.

Abstract

We propose a novel method for gradient-based optimization of black-box simulators using differentiable local surrogate models. In fields such as physics and engineering, many processes are modeled with non-differentiable simulators with intractable likelihoods. Optimization of these forward models is particularly challenging, especially when the simulator is stochastic. To address such cases, we introduce the use of deep generative models to iteratively approximate the simulator in local neighborhoods of the parameter space. We demonstrate that these local surrogates can be used to approximate the gradient of the simulator, and thus enable gradient-based optimization of simulator parameters. In cases where the dependence of the simulator on the parameter space is constrained to a low dimensional submanifold, we observe that our method attains minima faster than baseline methods, including Bayesian optimization, numerical optimization, and approaches using score function gradient estimators.