NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID: 1761 Individual Regret in Cooperative Nonstochastic Multi-Armed Bandits

### Reviewer 1

- It would be helpful to write down Exp3 more explicitly (at least using a numbered display) - it would be helpful to spell out the proposed cooperative MAB algorithm explicitly. It is somewhat hard to grasp how the current Alg. 1 and 4 relate to the proposed overall method. authors might wish to sharpen/reword the following: - "We can now obtain the same desired result.... " (what is the desired result?) - "The individual regret bound we introduced for the center-based..." (point to the bound e.g. using eqref) - ".and presented the center based cooperation policy." (point to the policy using eqref) - "This bound resolves an open question from [Cesa-Bianchi et al., 2019b] and also implies the result presented there" - "(i.e., agents are not partitioned to types.."