Exploring Deep-Reinforcement-Learning-Assisted Federated Learning for Online Resource Allocation in Privacy-Preserving EdgeIoT

Federated learning (FL) has been increasingly considered to preserve data training privacy from eavesdropping attacks in mobile-edge computing-based Internet of Things (EdgeIoT). On the one hand, accuracy FL can be improved by selecting IoT devices with large sets for training, which gives rise a higher energy consumption. other consumption reduced small FL, resulting falling accuracy. In this article, we formulate new resource allocation problem privacy-preserving EdgeIoT balance and device. We propose FL-enabled twin-delayed deep deterministic policy gradient (FL-DLT3) framework achieve optimal continuous domain. Furthermore, long short-term memory (LSTM) is leveraged FL-DLT3 predict time-varying network state while trained select allocate transmit power. Numerical results demonstrate that proposed achieves fast convergence (less than 100 iterations) accuracy-to-energy ratio 51.8% compared existing state-of-the-art benchmark.