Attraction and Repulsion: Unsupervised Domain Adaptive Graph Contrastive Learning Network
Attraction and Repulsion: Unsupervised Domain Adaptive Graph Contrastive Learning Network
复制标题
DOI：
10.1109/tetci.2022.3156044
复制
发表时间：
2022-10
影响因子：
5.3
通讯作者：
Man Wu;Shirui Pan;Xingquan Zhu
中科院分区：
文献类型：
--
作者：
Man Wu;Shirui Pan;Xingquan Zhu
图卷积网络（GCN）是与图数据相关的分析任务的重要技术。迄今为止，大多数 GCN 都是为单个图域设计的。由于图表示学习和跨图域的域适应的限制，它们无法将知识从不同的域（图）转移到不同的域（图）。本文提出了一种新颖的图对比学习网络（GCLN），用于无监督域自适应图学习。关键的创新是在每个单个图域内以及跨两个图域强制施加吸引力和排斥力。在每个图中，吸引力促使局部补丁节点特征与整个图的全局表示相似，而排斥力将排斥节点特征，以便它们可以将网络与其排列分开（即特定领域的图对比学习）。在两个图域中，吸引力鼓励两个域的节点特征在很大程度上保持一致，而排斥力则确保特征具有区分性以区分图域（即跨域图对比学习）。域内和跨域图对比学习是通过优化目标函数来进行的，该目标函数结合了源分类器和目标分类器损失、特定域对比损失和跨域对比损失。因此，使用图之间转移的知识可以促进图的特征学习。对真实世界数据集的实验表明，GCLN 的性能优于最先进的图神经网络算法。
Graph convolutional networks (GCNs) are important techniques for analytics tasks related to graph data. To date, most GCNs are designed for a single graph domain. They are incapable of transferring knowledge from/to different domains (graphs), due to the limitation in graph representation learning and domain adaptation across graph domains. This paper proposes a novel Graph Contrastive Learning Network (GCLN) for unsupervised domain adaptive graph learning. The key innovation is to enforce attraction and repulsion forces within each single graph domain, and across two graph domains. Within each graph, an attraction force encourages local patch node features to be similar to global representation of the entire graph, whereas a repulsion force will repel node features so they can separate network from its permutations (i.e. domain-specific graph contrastive learning). Across two graph domains, an attraction force encourages node features from two domains to be largely consistent, whereas a repulsion force ensures features are discriminative to differentiate graph domains (i.e. cross-domain graph contrastive learning). The within- and cross-domain graph contrastive learning is carried out by optimizing an objective function, which combines source classifier and target classifier loss, domain-specific contrastive loss, and cross-domain contrastive loss. As a result, feature learning from graphs is facilitated using knowledge transferred between graphs. Experiments on real-world datasets demonstrate that GCLN outperforms state-of-the-art graph neural network algorithms.