Hierarchical Graph Representation Learning with Differentiable Pooling
This addresses the problem of learning hierarchical graph representations for graph classification tasks, which is an incremental but important advancement for graph neural network research.
The paper tackles the limitation of flat graph neural networks (GNNs) for graph classification by proposing DiffPool, a differentiable graph pooling module that generates hierarchical representations, resulting in an average 5-10% accuracy improvement on benchmarks and achieving state-of-the-art on four out of five datasets.
Recently, graph neural networks (GNNs) have revolutionized the field of graph representation learning through effectively learned node embeddings, and achieved state-of-the-art results in tasks such as node classification and link prediction. However, current GNN methods are inherently flat and do not learn hierarchical representations of graphs---a limitation that is especially problematic for the task of graph classification, where the goal is to predict the label associated with an entire graph. Here we propose DiffPool, a differentiable graph pooling module that can generate hierarchical representations of graphs and can be combined with various graph neural network architectures in an end-to-end fashion. DiffPool learns a differentiable soft cluster assignment for nodes at each layer of a deep GNN, mapping nodes to a set of clusters, which then form the coarsened input for the next GNN layer. Our experimental results show that combining existing GNN methods with DiffPool yields an average improvement of 5-10% accuracy on graph classification benchmarks, compared to all existing pooling approaches, achieving a new state-of-the-art on four out of five benchmark data sets.