Pure Exploration in Asynchronous Federated Bandits
This work addresses robustness against latency and agent unavailability in federated bandits, which is an incremental improvement for distributed learning systems.
The paper tackles the problem of federated pure exploration in multi-armed and linear bandits, where multiple agents cooperatively identify the best arm under asynchronous conditions, achieving near-optimal sample complexities and efficient communication costs.
We study the federated pure exploration problem of multi-armed bandits and linear bandits, where $M$ agents cooperatively identify the best arm via communicating with the central server. To enhance the robustness against latency and unavailability of agents that are common in practice, we propose the first federated asynchronous multi-armed bandit and linear bandit algorithms for pure exploration with fixed confidence. Our theoretical analysis shows the proposed algorithms achieve near-optimal sample complexities and efficient communication costs in a fully asynchronous environment. Moreover, experimental results based on synthetic and real-world data empirically elucidate the effectiveness and communication cost-efficiency of the proposed algorithms.