I am currently a Ph.D. student at Sun Yat-sen University. My research interests include:
- (Reinforcement learning for) LLM, LLM RAG.
- Reinforcement learning for LLM agents.
- Open-world Graph Learning: open-set recognition(OSR), generalized category discovery (GCD), etc.
- (Graph) Contrastive Learning
- Robust Graph Learning: adversarial robustness, anomaly detection, etc.
I am always open to collaborations, and if you are interested in my research, feel free to contact me via email.
🔥 News
-
2025.05: 🎉🎉 Two papers are accepted by ICML 2025.
- 2025.01: 🎉🎉 One paper is accepted by ICLR 2025.
- 2024.12: Invited to be a reviewer for ICML 2025.
- 2024.12: 🎉🎉 The graph contrastive learning framework THESAURUS is accepted by AAAI 2025 (CCF A, Core A*).
- 2024.07: 🎉🎉 The mutual GNN-MLP distillation framework PROSPECT is accepted by CIKM 2024 (CCF B, Core A).
📝 Publications
Towards Understanding Parametric Generalized Category Discovery on Graphs
Bowen Deng, Lele Fu, Jialong Chen, Sheng Huang, Tianchi Liao, Tao Zhang, Chuan Chen.
Proceedings of the 42nd International Conference on Machine Learning (ICML) , 2025.
TL;DR: We provide the first rigorous theoretical answer to the core GCD question: “When and how do old classes help (parametric) generalized category discovery (on graphs)?”
THESAURUS: Contrastive Graph Clustering by Swapping Fused Gromov-Wasserstein Couplings
Bowen Deng, Tong Wang, Lele Fu, Sheng Huang, Chuan Chen, and Tao Zhang.
Proceedings of the AAAI Conference on Artificial Intelligence (AAAI ’25), February 25 – March 4, 2025, Philadelphia, Pennsylvania, USA.
TL;DR: We propose a novel graph contrastive learning framework based on Fused Gromov-Wasserstein optimal transport on graphs.
Mutual GNN-MLP Distillation for Robust Graph Adversarial Defense
Bowen Deng, Jialong Chen, Yanming Hu, Chuan Chen, Tao Zhang
Neural Networks (CCF-B), 2024
TL;DR: We prove the robustness and heterophily adaptability of Mutual GNN-MLP Distillation (MGMD) and further discuss its robustness against adaptive attacks.
PROSPECT: Learn MLPs on Graphs Robust against Adversarial Structure Attacks
Bowen Deng, Jialong Chen, Yanming Hu, Zhiyong Xu, Chuan Chen, Tao Zhang
Proceedings of the 33rd ACM International Conference on Information and Knowledge Management (CIKM ’24), October 21–25, 2024, Boise, ID, USA.
TL;DR: Tackle heteropihly, robustness, and inference scalability challenges with GNN-to-MLP knowledge distillation.
Dynamic Graph Convolutional Network: A Topology Optimization Perspective
Bowen Deng, Aimin Jiang
IEEE 31st International Workshop on Machine Learning for Signal Processing (MLSP), Gold Coast, Australia, 2021.
TL;DR: Refine the graph structure with pseudo labels with learnable edge scores.
-
Less is More: Federated Graph Learning with Alleviating Topology Heterogeneity from A Causal Perspective, Lele Fu, Bowen Deng , Sheng Huang, Tianchi Liao, Shirui Pan, Chuan Chen, ICML 2025
-
Graph Neural Ricci Flow: Evolving Feature from a Curvature Perspective, Jialong Chen, Bowen Deng, Zhen WANG, Chuan Chen, Zibin Zheng, ICLR 2025
-
Decoupling anomaly discrimination and representation learning: self-supervised learning for anomaly detection on attributed graph, Yanming Hu, Chuan Chen, Bowen Deng, Yujing Lai, Hao Lin, Zibin Zheng, Jing Bian, Data Science and Engineering, 2024
-
ADMM-Based TDOA Estimation, Yanping Zhu , Bowen Deng, Aimin Jiang, Xiaofeng Liu, IEEE Communications Letters, 2018
📖 Educations
- 2022.09 - present, Ph.D, Sun Yat-sen University. (Average: 88.95/100)
- 2019.06 - 2022.06, Master, Hohai University (Postgraduate Recommendation). (Average: 83.67/100)
- 2015.09 - 2019.06, Undergraduate, Hohai University. (Average: 87.74/100, Transcript En/Zh)
💻 Internships
Efficient Graph Signal Sampling and Reconstruction
Mentor: Prof. Antonio Ortega (University of Southern California)
When and where: 2021.07 - 2021.09, remote
- Developed a PyTorch-based package for graph signal sampling and reconstruction.
- Surveyed and implemented the most efficient existing algorithms for graph signal sampling and reconstruction.
- Proposed a data-driven graph signal sampling algorithm for complex signals mixup by low- and high-frequency signals.