GraphDF: A Discrete Flow Model for Molecular Graph Generation
This work addresses molecular graph generation for drug discovery, offering a novel approach that improves accuracy and efficiency over continuous methods.
The paper tackled the problem of molecular graph generation by proposing GraphDF, a discrete latent variable model based on normalizing flows, which outperformed prior methods on tasks like random generation and property optimization.
We consider the problem of molecular graph generation using deep models. While graphs are discrete, most existing methods use continuous latent variables, resulting in inaccurate modeling of discrete graph structures. In this work, we propose GraphDF, a novel discrete latent variable model for molecular graph generation based on normalizing flow methods. GraphDF uses invertible modulo shift transforms to map discrete latent variables to graph nodes and edges. We show that the use of discrete latent variables reduces computational costs and eliminates the negative effect of dequantization. Comprehensive experimental results show that GraphDF outperforms prior methods on random generation, property optimization, and constrained optimization tasks.