Constrained Bayesian Optimization for Automatic Chemical Design

TL;DR

This paper improves automatic chemical design by reformulating Bayesian optimization as a constrained problem, significantly increasing the validity of generated molecules by addressing the issue of querying latent points far from training data.

Contribution

It introduces a constrained Bayesian optimization approach to mitigate invalid molecule generation in variational autoencoder-based chemical design.

Findings

Conventional Bayesian optimization tends to produce invalid molecules.

Reformulating as a constrained problem improves molecule validity.

The approach addresses training set mismatch in generative tasks.

Abstract

Automatic Chemical Design is a framework for generating novel molecules with optimized properties. The original scheme, featuring Bayesian optimization over the latent space of a variational autoencoder, suffers from the pathology that it tends to produce invalid molecular structures. First, we demonstrate empirically that this pathology arises when the Bayesian optimization scheme queries latent points far away from the data on which the variational autoencoder has been trained. Secondly, by reformulating the search procedure as a constrained Bayesian optimization problem, we show that the effects of this pathology can be mitigated, yielding marked improvements in the validity of the generated molecules. We posit that constrained Bayesian optimization is a good approach for solving this class of training set mismatch in many generative tasks involving Bayesian optimization over the latent space of a variational autoencoder.