Evrard Garcelon
Papers
1
Total Citations
12
H-Index
1
About
Evrard Garcelon is a leading researcher in the field of online learning and bandit algorithms, with a particular focus on developing safe, conservative exploration strategies for real-world deployment. His work addresses a critical challenge: how to improve a suboptimal but reliable baseline policy—common in digital marketing, healthcare, finance, and robotics—without incurring unacceptable performance drops during learning. Garcelon’s most influential contribution, the paper "Improved Algorithms for Conservative Exploration in Bandits" (2020, 12 citations), provides theoretically grounded algorithms that guarantee the new policy never performs significantly worse than the existing baseline, even during the exploration phase. This breakthrough bridges the gap between theoretical bandit research and practical, risk-sensitive applications. By formalizing the trade-off between learning and safety, Garcelon has paved the way for deploying adaptive systems in high-stakes environments where mistakes are costly. His work is essential reading for anyone interested in responsible AI deployment, offering both rigorous proofs and actionable insights for practitioners.
Research Focus
Key Achievements
Top Papers
- 1Improved Algorithms for Conservative Exploration in Bandits12 citations · 2020