Random Walk Conformer: Learning Graph Representation from Long and Short Range

While graph neural networks (GNNs) have achieved notable success in various mining tasks, conventional GNNs only model the pairwise correlation 1-hop neighbors without considering long-term relations and high-order patterns, thus limiting their performances. Recently, several works addressed these issues by exploring motif, i.e., frequent subgraphs. However, methods usually require an unacceptable computational time to enumerate all possible combinations of motifs. In this paper, we introduce a new GNN framework, namely Random Walk Conformer (RWC), exploit global correlations local patterns based on random walk, which is promising method discover structure. Besides, propose walk encoding help RWC capture topological information, proven more expressive than spatial encoding. Extensive experiment results manifest that achieves state-of-the-art performance classification regression tasks. The source code available at https://github.com/b05901024/RandomWalkConformer.