Co-attention Graph Pooling for Efficient Pairwise Graph Interaction Learning

Graph Neural Networks (GNNs) have proven to be effective in processing and learning from graph-structured data. However, previous works mainly focused on understanding single graph inputs while many real-world applications require pair-wise analysis for data (e.g., scene matching, code searching, drug-drug interaction prediction). To this end, recent shifted their focus the between pairs of graphs. Despite improved performance, these were still limited that interactions considered at node-level, resulting high computational costs suboptimal performance. address issue, we propose a novel efficient graph-level approach extracting representations using co-attention pooling. Our method, Co-Attention Pooling (CAGPool), exhibits competitive performance relative existing methods both classification regression tasks datasets, maintaining lower complexity.