Decoding Molecular Graph Embeddings with Reinforcement Learning

TL;DR

This paper introduces RL-VAE, a reinforcement learning-based graph autoencoder that efficiently decodes molecular graphs from latent embeddings, simplifying the process compared to previous complex methods.

Contribution

The paper proposes a novel, simplified graph generator within a variational autoencoder framework that improves efficiency in molecular graph decoding.

Findings

Efficient decoding of molecular graphs achieved.

Simplified training process without complex decoders.

Potential for scalable molecular generation.

Abstract

We present RL-VAE, a graph-to-graph variational autoencoder that uses reinforcement learning to decode molecular graphs from latent embeddings. Methods have been described previously for graph-to-graph autoencoding, but these approaches require sophisticated decoders that increase the complexity of training and evaluation (such as requiring parallel encoders and decoders or non-trivial graph matching). Here, we repurpose a simple graph generator to enable efficient decoding and generation of molecular graphs.