Graph Deconvolutional Generation
This work addresses graph generation, a key task in science and engineering, by enhancing an existing model for specific applications like molecule generation, making it incremental in nature.
The authors tackled the problem of graph generation by improving the graph variational autoencoder (GVAE) model, which struggles with matching training distributions and requires expensive graph matching, by integrating a message passing neural network into its encoder and decoder, and demonstrated it on generating small organic molecules.
Graph generation is an extremely important task, as graphs are found throughout different areas of science and engineering. In this work, we focus on the modern equivalent of the Erdos-Renyi random graph model: the graph variational autoencoder (GVAE). This model assumes edges and nodes are independent in order to generate entire graphs at a time using a multi-layer perceptron decoder. As a result of these assumptions, GVAE has difficulty matching the training distribution and relies on an expensive graph matching procedure. We improve this class of models by building a message passing neural network into GVAE's encoder and decoder. We demonstrate our model on the specific task of generating small organic molecules