Abstract:
The goal of this interdisciplinary project is to enhance baseline models and methods for generative molecular modeling, aiming to generate novel molecules and perform inverse design to obtain molecular structures with desired properties. This project builds upon a previous interdisciplinary project (IDP), where a new generative model—Graph Variational Autoencoder (GVAE)—and a Mixture Model (MM) were developed for inverse design.
This project focuses on experiments and improvements involving the Junction Tree Variational Autoencoder (JTVAE), property prediction, and inverse design. We compare the results of property prediction, inverse design, and molecule generation with baseline models, previous models from the earlier IDP project, the SMILES-based variational autoencoder, and the newly proposed models.