Collaborative Top Distribution Identifications with Limited Interaction

Explore a 25-minute IEEE conference talk on collaborative learning strategies for Multi-Armed Bandits (MAB) problems, focusing on top-m distribution identifications with limited interaction. Delve into the intricacies of the Top-m problem in MAB, examining the collaborative learning approach and its algorithm. Understand the key differences between Top-1 and Top-m scenarios, and learn about the reduction technique and error amplification in this context. Presented by Nikolai Karpov, Qin Zhang, and Yuan Zhou from Indiana University at Bloomington and the University of Illinois at Urbana-Champaign, this talk provides valuable insights into advanced MAB problem-solving techniques.

Intro
Top-m problem in Multi-Armed Bandits (MAB)
Collaborative Learning for MAB
Algorithm
Difference between Top-1 and Top-m Top-m
Reduction
Error amplification