Dist-Hedge: A partial information setting based distributed non-stochastic sequence prediction algorithm

摘要

This paper focuses on the problem of distributed sequence prediction in a network of sparsely interconnected forecasting agents, where agents collaborate to achieve provably reasonable predictive performance. An expert assisted online learning algorithm Dist-Hedge of the consensus+innovations form is proposed, in which the agents aggregate experts' predictions by simultaneously processing the latest network losses (innovations) and the cumulative losses obtained from neighboring agents (consensus). This paper characterizes the sublinear regret of the agents' prediction performance with respect to the best forecasting expert in terms of network connectivity.