graphs

Likelihood-Free Inference and Generation of Molecular Graphs

Recent methods for generating novel molecules use graph representations of molecules and employ various forms of graph convolutional neural networks for inference. However, training requires solving an expensive graph isomorphism problem, which …