Tony Liu

liutony@seas.upenn.edu
Richards Building 4th Floor, Hamilton Walk
Google Scholar
About Me
Hello! I’m a PhD student in the Department of Computer Information Science at the University of Pennsylvania, jointly advised by Konrad Kording and Lyle Ungar. Prior to UPenn, I graduated from Williams College in 2016 with a degree in Computer Science and spent two years at IBM Watson Health working as a software engineer.
I’m interested in the intersection of causal inference and machine learning, specifically in applications to healthcare. I also dabble in all things CS.
Publications
- Liu T, Ungar LH, Kording KP. Quantifying causality in data science using quasi-experiments. Nature Computational Science. 2021. [DOI]
- Shen H, Liu T, Cui J, Borole P, Benjamin A, Kording K, Issadore D. A web-based automated machine learning platform to analyze liquid biopsy data. Lab on a Chip. 2020. [DOI]
- Liu T, Nicholas J, Theilig MM, Guntuku SC, Kording K, Mohr DC, Ungar LH. Machine Learning for Phone-Based Relationship Estimation: The Need to Consider Population Heterogeneity. Proceedings of the ACM on interactive, mobile, wearable and ubiquitous technologies. 2019 Dec 11. [DOI]
Workshop Papers
- Liu T, Ungar LH, Kording KP. Data-Driven Exclusion Criteria for Instrumental Variables. To appear, ICML Workshop on Neglected Assumptions in Causal Inference. 2021.
- Liu T, Ungar LH. Towards Cotenable and Causal Shapley Feature Explanations. AAAI Workshop on Trustworthy AI in Healthcare. 2021. [Paper] [Poster]
Preprints and submissions
- Liu T, Meyerhoff J, Eichstaedt J, Karr CJ, Kaiser SM, Kording KP, Mohr DC, Ungar LH. Predicting depression severity from text message sentiment. Under submission.
- Meyerhoff J, Liu T, Kording KP, Ungar LH, Kaiser SM, Karr CJ, Mohr DC. Prediction of change in depression, anxiety, and social anxiety using smartphone sensor features: longitudinal cohort study. Under submission.
- Liu T, Meyerhoff J, Mohr DC, Ungar LH, Kording KP. COVID-19 pandemic: every day feels like a weekday to most. medRxiv. 2020. [DOI]
Current Projects
- Automatic regression discontinuity design discovery
- Optimal exclusion criteria for observational studies
- Cotenability and causality in SHAP values [PDF]
- Overfitting detection using unlabelled data [PDF]
Posters and Presentations
- Liu T, Lawlor P, Ungar LH, Kording KP. Evaluating clinical guidelines using regression discontinuity design. AMIA 2020 Annual Symposium. [PDF]
Teaching