Juliette Achddou (CRIStAL) : Two shades of distributed online convex optimization

Distributed online convex optimization (DOCO) has emerged as a fundamental framework for modeling sequential decision-making in decentralized environments, particularly when data is distributed across multiple agents with privacy considerations. The key challenge in this setting is to understand how regret depends jointly on the time horizon and the topology of the communication network connecting the agents. I will try to investigate two important extensions of the classical DOCO framework: delayed feedback, and random agent availability, both of which capture realistic constraints in distributed systems.