Robust Asynchronous and Network-Independent Cooperative Learning

TL;DR

This paper introduces a robust cooperative learning method for distributed networks that functions effectively despite asynchronous communication, message delays, losses, and directed links, ensuring all agents converge on the correct hypothesis.

Contribution

It proposes a new asynchronous, network-independent learning rule that guarantees convergence to the optimal hypothesis under various network conditions.

Findings

Beliefs on incorrect hypotheses decay exponentially over time.

The method is robust to message delays and losses.

Numerical experiments confirm effectiveness across different network setups.

Abstract

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.