Heavy-Tailed Noise in Linear and Kernel Bandits

Explore the challenges of heavy-tailed noise in linear and kernel bandit algorithms through this 14-minute workshop presentation by Artin Tajdini from the Paul G. Allen School of Computer Science & Engineering. Delve into the theoretical and practical implications of non-Gaussian noise distributions in multi-armed bandit problems, examining how heavy-tailed noise affects the performance and convergence properties of both linear and kernel-based bandit algorithms. Learn about the mathematical frameworks used to analyze these scenarios and discover potential solutions for handling the increased variance and outliers that characterize heavy-tailed distributions in online learning environments.