(GAAN)Graph Adaptive Attention Network
Keywords: Keywords: Non-Euclidean, GCN, Adaptive Attention Mechanism, Multi-Head Graph Convolution
TL;DR: We have proposed AAM and MHGC to construct GCNA, which solves the differences between neighbor nodes to the central node.
Abstract: Non-Euclidean data, such as social networks, and citation relationships between documents, has node information and structural information. Graph Convolutional Network(GCN) can automatically learn node features and association information between nodes. For example, social networks are very suitable for using graph data to express, such as nodes in social networks and the relationship between nodes. Users (with ID information, etc.), posts are nodes, the relationship between user A and user B is attention, and the relationship between users and posts may be published or forwarded. Through such a graph, it is possible to analyze who and what users are interested in, and further generate the recommendation system. The core ideology of the graph convolutional network is to aggregate node information by using edge information, thereby generating a new node feather. Considering the numbers and different contributions of neighbor nodes to the central node, we design the Adaptive Attention Mechanism(AAM). To further enhance the representational capability of the model, we utilize Multi-Head Graph Convolution(MHGC). Based on AAM and MHGC, we contrive the novel Graph Adaptive Attention Network (GAAN). Experiments on the CORA dataset show that the classification accuracy has achieved 85.6%.
Submission Number: 1
Loading