Better Algorithms for Stochastic Bandits with Adversarial Corruptions
This addresses a robust decision-making problem in machine learning, offering a novel solution for handling adversarial corruptions in bandit settings.
The paper tackles the stochastic multi-armed bandits problem with adversarial corruption by presenting a new algorithm that achieves nearly optimal regret, substantially improving upon previous work, and can tolerate significant corruption with minimal performance degradation.
We study the stochastic multi-armed bandits problem in the presence of adversarial corruption. We present a new algorithm for this problem whose regret is nearly optimal, substantially improving upon previous work. Our algorithm is agnostic to the level of adversarial contamination and can tolerate a significant amount of corruption with virtually no degradation in performance.