Distributed Online Composite Optimization with Delayed Feedback

This paper focuses on distributed online composite optimization with feedback delays. Using the delayed feedback and signs of relative states, a distributed online proximal gradient algorithm is proposed and then the performance of this algorithm is evaluated by dynamic regret. Under basic assumptions, a bound on dynamic regret for each agent is established, which is consistent with the existing result in the setting without regularizers. In the end, numerical examples are presented to support the theoretical finding.