Taha Bahadori
Principal Scientist, Amazon Affiliated Faculty, UW Department of Electrical & Computer Engineering
Research Interests
My primary research area is the intersection of causal inference and deep learning. I apply causal inference techniques to understand large language models and improve their safety. I am also invested in using modern AI to solve the challenging problems in healthcare.
Selected Publications
Fast Training Dataset Attribution via In-Context Learning
With M. Fotouhi and others (2024)
ICML Workshop on In-Context Learning
Multiply-Robust Causal Change Attribution
with V. Quintas-Martinez and others (2024)
International Conference on Machine Learning (ICML)
End-to-End Balancing for Causal Continuous Treatment-Effect Estimation
With E. Tchetgen-Tchetgen and D. Heckerman (2022)
International Conference on Machine Learning (ICML)
Debiasing Concept-based Explanations with Causal Analysis
With D. Heckerman (2021)
International Conference on Learning Representations (ICLR)
RETAIN: Interpretable Predictive Model in Healthcare using Reverse Time Attention Mechanism
With SunLab Group (2016)
Conference on Neural Information Processing Systems (NeurIPS)
Doctor AI: Predicting Clinical Events via Recurrent Neural Networks
With SunLab Group (2016)
Machine Learning for Healthcare Conference (MLHC)
Education
- Ph.D., University of Southern California, 2015
- B.S., Sharif University of Technology, 2008
Contact
Email:
bahadori(at)uw.edu