This website contains information regarding the paper Modular Flows: Differential Molecular Generation.
TL;DR: We propose generative graph normalizing flow models, based on a system of coupled node ODEs, that repeatedly reconcile locally toward globally aligned densities for high quality molecular generation
A key challenge of molecular generative models is to be able to generate valid molecules, according to various criteria for molecular validity or feasibility. It is a common practice to use external chemical software as rejection oracles to reduce or exclude invalid molecules, or do validity checks as part of autoregressive generation [1,2,3] . An important open question has been whether generative models can learn to achieve high generative validity intrinsically, i.e., without being aided by oracles or performing additional checks.
