Robust Asynchronous and Network-Independent Cooperative Learning
This work addresses robust distributed learning for multi-agent systems in unreliable network environments, representing an incremental improvement over existing weak communication models.
The paper tackles the problem of cooperative learning in distributed networks with unreliable communication, proposing a robust learning rule that handles asynchrony, delays, message losses, and directed links, and shows that all agents' beliefs converge exponentially to the optimal hypothesis.
We consider the model of cooperative learning via distributed non-Bayesian learning, where a network of agents tries to jointly agree on a hypothesis that best described a sequence of locally available observations. Building upon recently proposed weak communication network models, we propose a robust cooperative learning rule that allows asynchronous communications, message delays, unpredictable message losses, and directed communication among nodes. We show that our proposed learning dynamics guarantee that all agents in the network will have an asymptotic exponential decay of their beliefs on the wrong hypothesis, indicating that the beliefs of all agents will concentrate on the optimal hypotheses. Numerical experiments provide evidence on a number of network setups.