Sparse Graph Attention Networks

Graph Neural Networks (GNNs) have proved to be an effective representation learning framework for graph-structured data, and have achieved state-of-the-art performance on many practical predictive tasks. Among the variants of GNNs, Graph Attention Networks (GATs) improve the performance of many graph learning tasks through a dense attention mechanism. However, real-world graphs are often very large and noisy, and GATs are prone to overfitting if not regularized properly. In this paper, we propose Sparse Graph Attention Networks (SGATs) that learn sparse attention coefficients under an L0-norm regularization, and the learned sparse attentions are then used for all GNN layers, resulting in an edge-sparsified graph. By doing so, we can identify noisy/task-irrelevant edges, and thus perform feature aggregation on most informative neighbors. Extensive experiments on synthetic and real-world (assortative and disassortative) graph learning benchmarks demonstrate the superior performance of SGATs. Furthermore, the removed edges can be interpreted intuitively and quantitatively. To the best of our knowledge, this is the first graph learning algorithm that shows significant redundancies in graphs and edge-sparsified graphs can achieve similar (on assortative graphs) or sometimes higher (on disassortative graphs) predictive performances than original graphs. Our code is available at https://github.com/Yangyeeee/SGAT.