Resource Management for Blockchain-enabled Federated Learning: A Deep\n Reinforcement Learning Approach

Abstract

Blockchain-enabled Federated Learning (BFL) enables mobile devices to\ncollaboratively train neural network models required by a Machine Learning\nModel Owner (MLMO) while keeping data on the mobile devices. Then, the model\nupdates are stored in the blockchain in a decentralized and reliable manner.\nHowever, the issue of BFL is that the mobile devices have energy and CPU\nconstraints that may reduce the system lifetime and training efficiency. The\nother issue is that the training latency may increase due to the blockchain\nmining process. To address these issues, the MLMO needs to (i) decide how much\ndata and energy that the mobile devices use for the training and (ii) determine\nthe block generation rate to minimize the system latency, energy consumption,\nand incentive cost while achieving the target accuracy for the model. Under the\nuncertainty of the BFL environment, it is challenging for the MLMO to determine\nthe optimal decisions. We propose to use the Deep Reinforcement Learning (DRL)\nto derive the optimal decisions for the MLMO.\n