Complete the Missing Half: Augmenting Aggregation Filtering with Diversification for Graph Convolutional Networks
share
The core operation of Graph Neural Networks (GNNs) is the aggregation enabled by the graph Laplacian or message passing, which filters the neighborhood information of the nodes. Though effective for various tasks, they are the potentially problematic factor of all GNN methods, as they force the node representations to be similar, making the nodes gradually lose their identity and become indistinguishable. In this paper, we augment the aggregation operations with their dual, i.e. diversification operators that make the node more distinct and preserve the identity. Such augmentation replaces the aggregation with a two-pass filtering process that, in theory, is beneficial for enriching the node representations. In practice, the two-pass filters can be easily patched on existing GNN methods with diverse training strategies, including spectral and spatial (message passing) methods. When patched on baselines, we observe the significant performance boost on 8 node and graph classification tasks.
READ FULL TEXT
