Improving black-box optimization in VAE latent space using decoder uncertainty

TL;DR

This paper introduces a method to improve black-box optimization in VAE latent spaces by using a robust estimator of decoder uncertainty, enhancing the validity and quality of generated objects without altering the model architecture.

Contribution

It proposes an importance sampling-based estimator for decoder epistemic uncertainty, guiding optimization more reliably in high-dimensional structured latent spaces.

Findings

Improved validity of generated samples in molecular design and digit generation.

Enhanced trade-off between optimizing black-box objectives and sample validity.

Method works without modifying existing VAE architectures or training procedures.

Abstract

Optimization in the latent space of variational autoencoders is a promising approach to generate high-dimensional discrete objects that maximize an expensive black-box property (e.g., drug-likeness in molecular generation, function approximation with arithmetic expressions). However, existing methods lack robustness as they may decide to explore areas of the latent space for which no data was available during training and where the decoder can be unreliable, leading to the generation of unrealistic or invalid objects. We propose to leverage the epistemic uncertainty of the decoder to guide the optimization process. This is not trivial though, as a naive estimation of uncertainty in the high-dimensional and structured settings we consider would result in high estimator variance. To solve this problem, we introduce an importance sampling-based estimator that provides more robust estimates of epistemic uncertainty. Our uncertainty-guided optimization approach does not require modifications of the model architecture nor the training process. It produces samples with a better trade-off between black-box objective and validity of the generated samples, sometimes improving both simultaneously. We illustrate these advantages across several experimental settings in digit generation, arithmetic expression approximation and molecule generation for drug design.