CLDG: Contrastive Learning on Dynamic Graphs
This work addresses a critical bottleneck in dynamic graph learning for researchers and practitioners, offering a more efficient and effective unsupervised approach, though it appears incremental as it builds on existing contrastive learning paradigms.
The paper tackles the problem of performance drop in unsupervised dynamic graph representation learning due to semantic changes during augmentation, by proposing CLDG, a framework that leverages temporal translation invariance to extract persistent signals. The method outperforms eight unsupervised baselines and shows competitive results against semi-supervised methods, while reducing model parameters by an average of 2,001.86 times and training time by 130.31 times across seven datasets.
The graph with complex annotations is the most potent data type, whose constantly evolving motivates further exploration of the unsupervised dynamic graph representation. One of the representative paradigms is graph contrastive learning. It constructs self-supervised signals by maximizing the mutual information between the statistic graph's augmentation views. However, the semantics and labels may change within the augmentation process, causing a significant performance drop in downstream tasks. This drawback becomes greatly magnified on dynamic graphs. To address this problem, we designed a simple yet effective framework named CLDG. Firstly, we elaborate that dynamic graphs have temporal translation invariance at different levels. Then, we proposed a sampling layer to extract the temporally-persistent signals. It will encourage the node to maintain consistent local and global representations, i.e., temporal translation invariance under the timespan views. The extensive experiments demonstrate the effectiveness and efficiency of the method on seven datasets by outperforming eight unsupervised state-of-the-art baselines and showing competitiveness against four semi-supervised methods. Compared with the existing dynamic graph method, the number of model parameters and training time is reduced by an average of 2,001.86 times and 130.31 times on seven datasets, respectively.