Yi Su
印Postdoctoral Researcher
Berkeley Artificial Intelligence Research Laboratory
University of California, Berkeley
Email: ys756@cornell.edu, suyi@berkeley.edu
About Me
I am a PostDoctoral Researcher at EECS, UC Berkeley, working with Professor Sergey Levine within the BAIR lab. Previously, I obtained my PhD in Statistics from Cornell University, advised by Professor Thorsten Joachims. Prior to that, I received my BSc (Honors) in Mathematics from Nanyang Technological University in the beautiful Singapore. During my undergraduate years, I spent a summer at University of California, Berkeley and a year at the Department of Statistics and Applied Probability, National University of Singapore. In Spring 2019, I worked as a Research Intern at the Machine Learning Group at Microsoft Research NYC, advised by Miro Dudik and Akshay Krishnamurthy. In Summer 2020, I worked as a Research Intern at Bloomberg AI, advised by Anju Kambadur.
I am interested in machine learning methods, algorithms, and systems. Specifically I am interested in learning from user behavioral data and implicit feedback in search engines, recommender systems and multi-sided market platforms. My current interest lies in off-policy evaluation and learning in contextual bandits and reinforcement learning.
Preprints and Publications
- Optimizing Rankings for Recommendation in Matching MarketsYi Su, Magd Bayoumi, Thorsten Joachims[PDF] World Wide Web Conference (WWW), 2022
- Online Adaptation to Label Distribution ShiftRuihan Wu, Chuan Guo, Yi Su, Kilian Q Weinberger[PDF] Neural Information Processing Systems (NeurIPS), 2021
- Recommendations as TreatmentsThorsten Joachims, Ben London, Yi Su, Adith Swaminathan, Lequn Wang(to appear in) A.I. Magazine, 2021
- Doubly robust off-policy evaluation with shrinkageYi Su, Maria Dimakopoulou, Akshay Krishnamurthy, Miroslav Dudik[PDF] International Conference on Machine Learning (ICML), 2020
- Adaptive estimator selection for off-policy evaluationYi Su, Pavithra Srinath, Akshay Krishnamurthy[PDF] International Conference on Machine Learning (ICML), 2020
- Off-policy Bandits with Deficient SupportNoveen Sachdeva*, Yi Su*, Thorsten Joachims[PDF] ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD), 2020
- CAB: Continuous Adaptive Blending for Policy Evaluation and LearningYi Su*, Lequn Wang*, Michele Santacatterina, Thorsten Joachims[PDF] International Conference on Machine Learning (ICML), 2019
- Learning from Logged Bandit Feedback of Multiple LoggersYi Su, Aman Agarwal, Thorsten Joachims[PDF] CausalML Workshop at the International Conference on Machine Learning (CausalML), 2018
Talks
- Off-policy Evaluation and Learning for Interactive SystemsInvited talk at SIGIR'21 Workshop on Causality Search and Recommendation, June 2021.Invited talk at SIGIR'21 Workshop on Deep Reinforcement Learning for Information Retrieval, June 2021.
- Adaptive Estimator Selection for Off-policy EvaluationNetflix Research Seminar, June 2021.RL Theory Virtual Seminar, March 2021.
- Off-policy Bandits with Deficient SupportBloomberg AI, August 2020.
Service
Conference Reviewing
- International Conference on Machine Learning (ICML), 2019, 2020 (top reviewer), 2021 (expert reviewer)
- Neural Information Processing Systems (NeurIPS), 2019, 2020, 2021
- Conference on Artificial Intelligence (AAAI), 2020, 2021
- International Conference on Learning Representations (ICLR), 2021
- NeurIPS Workshop: Offline Reinforcement Learning, 2020
Program Committee
- ICML Workshop: Theoretical Foundations of Reinforcement Learning, 2020