Pooling in Graph Convolutional Neural Networks
This work provides incremental insights into graph pooling for researchers in graph neural networks.
The paper empirically evaluated multiple graph pooling methods with three GCNN architectures, finding that DiffPool improves classification accuracy and TAGCN performs comparably or better than alternatives, especially on larger, sparser graphs.
Graph convolutional neural networks (GCNNs) are a powerful extension of deep learning techniques to graph-structured data problems. We empirically evaluate several pooling methods for GCNNs, and combinations of those graph pooling methods with three different architectures: GCN, TAGCN, and GraphSAGE. We confirm that graph pooling, especially DiffPool, improves classification accuracy on popular graph classification datasets and find that, on average, TAGCN achieves comparable or better accuracy than GCN and GraphSAGE, particularly for datasets with larger and sparser graph structures.