Cooperative Online Learning with Feedback Graphs
This work addresses communication and feedback interplay in multi-agent learning, but it is incremental as it extends existing theoretical frameworks.
The paper tackles the problem of cooperative online learning with feedback graphs by bounding network regret in terms of the independence number of the strong product between communication and feedback graphs, recovering known bounds and proving an instance-based lower bound.
We study the interplay between communication and feedback in a cooperative online learning setting, where a network of communicating agents learn a common sequential decision-making task through a feedback graph. We bound the network regret in terms of the independence number of the strong product between the communication network and the feedback graph. Our analysis recovers as special cases many previously known bounds for cooperative online learning with expert or bandit feedback. We also prove an instance-based lower bound, demonstrating that our positive results are not improvable except in pathological cases. Experiments on synthetic data confirm our theoretical findings.