An Unpooling Layer for Graph Generation

TL;DR

This paper introduces a trainable unpooling layer for graph generation that enlarges and refines graph structures, improving molecular generation tasks within GAN frameworks.

Contribution

It presents a novel trainable unpooling layer for graphs, applicable in autoencoders and GANs, with proven connectivity preservation and broad applicability to connected graphs.

Findings

Improved molecular generation on QM9 and ZINC datasets.

Unpooling layer outperforms adjacency-matrix-based methods.

Ensures connected graph structures after unpooling.

Abstract

We propose a novel and trainable graph unpooling layer for effective graph generation. Given a graph with features, the unpooling layer enlarges this graph and learns its desired new structure and features. Since this unpooling layer is trainable, it can be applied to graph generation either in the decoder of a variational autoencoder or in the generator of a generative adversarial network (GAN). We prove that the unpooled graph remains connected and any connected graph can be sequentially unpooled from a 3-nodes graph. We apply the unpooling layer within the GAN generator. Since the most studied instance of graph generation is molecular generation, we test our ideas in this context. Using the QM9 and ZINC datasets, we demonstrate the improvement obtained by using the unpooling layer instead of an adjacency-matrix-based approach.