2020/03/28 by Min Shi, Shi, Min, Yufei Tang +3
Computer Science · #Advanced Graph Neural Networks #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Social and Information Networks (cs.SI) #Topic Modeling #cs.SI
paper · pdf · doi:10.48550/arxiv.2003.12806
7 pages
arxiv created 2020/03/28 · openalex publication_date 2020/03/28 · arxiv updated 2020/03/31 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
In traditional Graph Neural Networks (GNN), graph convolutional learning is carried out through topology-driven recursive node content aggregation for network representation learning. In reality, network topology and node content are not always consistent because of irrelevant or missing links between nodes. A pure topology-driven feature aggregation approach between unaligned neighborhoods deteriorates learning for nodes with poor structure-content consistency, and incorrect messages could propagate over the whole network as a result. In this paper, we advocate co-alignment graph convolutional learning (CoGL), by aligning the topology and content networks to maximize consistency. Our theme is to force the topology network to respect underlying content network while simultaneously optimizing the content network to respect the topology for optimized representation learning. Given a network, CoGL first reconstructs a content network from node features then co-aligns the content network and the original network though a unified optimization goal with (1) minimized content loss, (2) minimized classification loss, and (3) minimized adversarial loss. Experiments on six benchmarks demonstrate that CoGL significantly outperforms existing state-of-the-art GNN models.