- Keywords: GNN, graph pooling, graph representation learning
- Abstract: Graphs are known to have complicated structures and have myriad applications. How to utilize deep learning methods for graph classification tasks has attracted considerable research attention in the past few years. Two properties of graph data have imposed significant challenges on existing graph learning techniques. (1) Diversity: each graph has a variable size of unordered nodes and diverse node/edge types. (2) Complexity: graphs have not only node/edge features but also complex topological features. These two properties motivate us to use multiplex structure to learn graph features in a diverse way. In this paper, we propose a simple but effective approach, MxPool, which concurrently uses multiple graph convolution networks and graph pooling networks to build hierarchical learning structure for graph representation learning tasks. Our experiments on numerous graph classification benchmarks show that our MxPool has marked superiority over other state-of-the-art graph representation learning methods. For example, MxPool achieves 92.1% accuracy on the D&D dataset while the second best method DiffPool only achieves 80.64% accuracy.
- Code: https://github.com/JucatL/MxPool