📝 Selected Publications

Causal Representation Learning from Multimodal Biomedical Observations

Yuewen Sun^*, Lingjing Kong^*, Guangyi Chen, Loka Li, Gongxu Luo, Zijian Li, Yixuan Zhang, Yujia Zheng, Mengyue Yang, Petar Stojanov, Eran Segal, Eric P. Xing, Kun Zhang (^*Equal contribution)

Project

• Theoretically, we establish identifiability guarantees for the latent components under multimodal observations. Empirically, we develop an estimation framework to recover the latent components within each modality

Identifying Latent State-Transition Processes for Individualized RL

Yuewen Sun, Biwei Huang, Yu Yao, Donghuo Zeng, Xinshuai Dong, Songyao Jin, Boyang Sun, Roberto Legaspi, Kazushi Ikeda, Peter Spirtes, Kun Zhang

Project

• We aim to identify the latent state-transition processes from observed state-action trajectories, facilitating the learning of personalized RL policies.
• Theoretical identifiability is guaranteed under both finite and infinite latent factor conditions, supporting the framework’s robustness.

ACAMDA: Improving Data Efficiency in Reinforcement Learning Through Guided Counterfactual Data Augmentation

Yuewen Sun, Erli Wang, Biwei Huang, Chaochao Lu, Lu Feng, Changyin Sun, Kun Zhang

Project

• We employ counterfactual reasoning to generate augmented datasets, enabling agents to make unbiased decisions, and model causal relationships within the system to ensure adaptability across heterogeneous environments.

Learning Temporally Causal Latent Processes from General Temporal Data

Weiran Yao^*, Yuewen Sun^*, Alex Ho, Changyin Sun, Kun Zhang

(^*Equal contribution)

Project

• We propose two provable conditions under which temporally causal latent processes can be identified from their observed nonlinear mixtures.
• We develop a theoretically-grounded training framework that enforces the assumed conditions through proper constraints.