REPUBLIC: Responsible AI with Reinforcement Learning

Our works “Interactive and Concentrated Differential Privacy for Bandits”, “Pure Exploration in Bandits with Linear Constraints” and “On the Complexity of Differentially Private Best-Arm Identification with Fixed Confidence” are accepted in EWRL (European Workshop on Reinforcement Learning), 2023.

Continue reading

We begin the REPUBLIC project supported by ANR (French National Research Agency) as a part of the JCJC (young researcher) program. In REPUBLIC, our goal is developing a unified framework to address robustness, privacy, and unbiasedness (fairness) in online learning as Reinforcement Learning with constraints, building the theory of Reinforcement…

Continue reading