<ul><li>This study focuses on multi-agent asynchronous online optimization with delays.</li><li>The existing regret bound for this problem is improved from O(sqrt(d*T)) to O(d*log(T)).</li><li>A delayed variant of the follow-the-leader algorithm (FTDL) is proposed to exploit the strong convexity of functions.</li><li>Experimental results show that the approximate FTDL performs better than the existing algorithm in the strongly convex case.</li></ul>

Revisiting Multi-Agent Asynchronous Online Optimization with Delays: the Strongly Convex Case

Discover more