Multi-Agent Deep Learning for Network Control

Earlier at Télécom SudParis, I proposed multi-agent deep learning-based solutions for load balancing and queue management in networks. This includes graph convolutional reinforcement learning for collaborative queuing agents, where network entities cooperate to route and schedule traffic, and deep reinforcement learning for smart queue management.

I also studied edge intelligence in IoT: non-cooperative game-theoretic models for edge server selection, including in federated learning settings, where IoT devices autonomously choose edge servers to reduce both learning error and communication cost.