Selected Research Projects

Deep Graph Contrastive Learning

Graph Contrastive Learning with Adaptive Augmentation

Existing deep Graph Neural Networks (GNNs) require abundant labeled data for training, which are often time-consuming and labor-intensive to collect. Can we learn generic knowledge directly from massive amounts of unlabeled data, considering that unlabeled data are accessible in most practical scenarios? Our research answers this question through the lens of graph contrastive learning, which trains the GNN model by maximizing consistency between augmented versions of given graphs. Our work significantly improves previous state-of-the-art unsupervised graph learning performance, sometimes even outperforming several supervised counterparts.

Further reading:

  • Continue reading our blog post on deep graph contrastive learning
  • Read our two research papers on generic graph contrastive learning, GRACE (GRL+@ICML 2020) and GCA (WWW 2021)
  • Read our empirical study of graph contrastive learning (NeurIPS 2021, Dataset and Benchmark track)
  • Read our paper on heterogeneous graph contrastive learning (SDM 2022)
  • Check out our maintained collection on self-supervised graph representation learning resources
  • Have a try at our developed PyGCL, a PyTorch library that features modularized components, standardized evaluation, and experiment management for graph contrastive learning

Recommendation with Graph Neural Networks

Session-based Recommendation with Graph Neural Networks

Nowadays, recommender systems have become an indispensable tool for efficient and effective decision making in an information overloaded era. Considering that users, items, and contexts within recommender systems are tightly connected, a natural question is whether graph structures can be helpful for mining users’ complex behaviors to enhance recommendations. Our work in this rigorous field provides affirmative answers to this question. Particularly, we develop the very first graph-based recommender systems for sequential session data as well as other recommendation settings such as multimedia recommendation.

Further reading:

  • Continue reading our blog post on session-based recommendation with graph neural networks
  • Read our two session-based recommendation models with graph neural networks, SR-GNN (AAAI 2019) and TAGNN (SIGIR 2020)
  • Read our work on graph structure learning for multimedia recommendation, LATTICE (ACMMM 2021) and its extended model MICRO (TKDE Major Revision)