Efficient Learning of Non-Autoregressive Graph Variational Autoencoders for Molecular Graph Gneration
Journal
J. Cheminformatics(Journal of Cheminformatics)
Date
2019.11.21
Abstract
With the advances in deep learning, deep generative models combined with graph neural networks have been successfully applied to data-driven molecular graph generation. Early work based on the non-autoregressive approach has been effective in generating molecular graphs fast and efficiently, but has suffered from low performance. In this paper, we present an improved learning method of a graph variational autoencoder for efficient molecular graph generation in a non-autoregressive manner. We introduce the three additional learning objectives and incorporate them into the training of the model: approximate graph matching, reinforcement learning, and auxiliary property prediction. We demonstrate the effectiveness of the proposed method by evaluating it on molecular generation tasks with QM9 and ZINC datasets. The model generates molecular graphs with high chemical validity and diversity compared to existing non-autoregressive methods. It can also conditionally generate molecular graphs satisfying various target conditions.
