Multi-Order-Content-Based Adaptive Graph Attention Network for Graph Node Classification

In graph-structured data, the node content contains rich information. Therefore, how to effectively utilize is crucial improve performance of graph convolutional networks (GCNs) on various analytical tasks. However, current GCNs do not fully content, especially multi-order content. For example, attention (GATs) only focus low-order while high-order completely ignored. To address this issue, we propose a novel network with adaptability that could features Its core idea has following novelties: First, constructed mechanism evaluate weights. Second, can i.e., it combines mechanisms high- and Furthermore, adaptability, perform good trade-off between according task requirements. Lastly, applied constructing structural symmetry. This more reasonably weights nodes, thereby improving convergence network. addition, conducted experiments multiple datasets compared proposed model state-of-the-art models in dimensions. The results validate feasibility effectiveness model.