Hengrui Cai

“Our research thrives on the era of causal evolution with the new generation of artificial intelligence, to establish reliable, powerful, and interpretable solutions to real-world problems.”

Solving Real-World Problems

Professor Hengrui Cai studies methodology and theory in causal inference, reinforcement learning, and graphical models and their interchanges as she works to develop reliable and interpretable solutions to real-world problems. Currently, her focus is individualized optimal decision making — an artificial intelligence paradigm tailored to an individual’s characteristics. “My work involves complex data, policy optimization and evaluation in reinforcement learning, and causal discovery for high-dimensional individual mediation analysis,” she says. “This work is motivated by fields such as precision medicine, customized economics, modern epidemiology and personalized marketing.”

Optimizing Individualized Treatment

Leveraging cutting-edge techniques in causal inference with reinforcement learning, Professor Cai aims to address certain issues with electronic health records, such as the need to coordinate multiple datasets from heterogeneous studies, utilize incomplete data structures, address continuous treatment domains, and be robust to unmeasured confounding variables. In particular, she is focused on patients in intensive care units. “I am working to develop an optimal individualized treatment rule, stored in electronic health records, for ICU patients,” she says. “Developing an individualized treatment rule can help assign the right treatment to the right patients at the right time.”

Providing Next-Generation Analysis

Another area of interest for Professor Cai is discovering causality among variables. For example, with COVID-19, quantifying the causal effects of approaches aimed at reducing the spread of the virus, regulated by other mediators, is challenging. “My team and I integrate state-of-the-art approaches in causal inference and graphical models to establish a new statistical framework to comprehensively characterize different sources of causal effects through individual mediators,” she explains. “The goal is the next-generation analysis of causal effects, interpreting the causal mechanism contributing at an individual level and identifying the crucial component in various processes, such as virus spread, gene expression or supply chains.

Education

Ph.D., Statistics, North Carolina State University

Select Publications

• Cai, H. *, Shi, C. *, Song, R., & Lu, W. (2023). Jump Q-Learning for Individualized Decision Making with Continuous Treatments. Journal of Machine Learning Research, 24(140), 1-92.
• Zhang, W., Wu, T., Wang, Y., Cai, Y., & Cai, H. (2023) Towards Trustworthy Explanation: On Causal Rationalization. In International Conference on Machine Learning (ICML 2023).
• Watson, RA. *, Cai, H. *, An, X., McLean, S., & Song, R. (2023). On Heterogeneous Treatment Effects in Heterogeneous Causal Graphs. In International Conference on Machine Learning (ICML 2023).
• Cai, H., Lu, W., Marceau West R., Mehrotra DV., & Huang, L. (2022). CAPITAL: Optimal Subgroup Identification via Constrained Policy Tree Search. Statistics in Medicine. 2022;1-14. DOI:10.1002/sim.9507.
• Cai, H. *, Shi, C *., Song, R., & Lu, W. (2021). Deep Jump Learning for Off-Policy Evaluation in Continuous Treatment Settings. Advances in Neural Information Processing Systems (NeurIPS), 34, 15285-15300.
• Cai, H., Song, R., & Lu, W. (2021). ANOCE: Analysis of Causal Effects with Multiple Mediators via Constrained Structural Learning. In International Conference on Learning Representations (ICLR).
• Cai, H., Song, R., & Lu, W. (2021). GEAR: On Optimal Decision Making with Auxiliary Data. Stat, 10(1):e399.
• Cai, H., Lu, W., & Song, R. (2020). On Validation and Planning of An Optimal Decision Rule with Application in Healthcare Studies. In International Conference on Machine Learning (ICML) (pp. 1262-1270). PMLR.