Joint Optimization of Energy Consumption and Completion Time in Federated Learning

Abstract

Federated Learning (FL) is an intriguing distributed machine learning approach due to its privacy-preserving characteristics. To balance the trade-off between energy and execution latency, and thus accommodate different demands and application scenarios, we formulate an optimization problem to minimize a weighted sum of total energy consumption and completion time through two weight parameters. The optimization variables include bandwidth, transmission power and CPU frequency of each device in the FL system, where all devices are linked to a base station and train a global model collaboratively. Through decomposing the non-convex optimization problem into two subproblems, we devise a resource allocation algorithm to determine the bandwidth allocation, transmission power, and CPU frequency for each participating device. We further present the convergence analysis and computational complexity of the proposed algorithm. Numerical results show that our proposed algorithm not only has better performance at different weight parameters (i.e., different demands) but also outperforms the state of the art.

Resource Allocation of Federated Learning for the Metaverse with Mobile Augmented Reality

“…Applying FL to MAR applications for the Metaverse still has high requirements of device memory, computational capability and communication bandwidth. There are several challenges to the deployment of FL to MAR applications: (1) Limited bandwidth results in long latency between clients and the server and, consequently, affects the convergence time of the global model. (2) A large amount of energy is needed because a satisfactory model requires quite a few local computation and communication rounds.…”

“…Comparison with our ICDCS paper [1]. We extend the algorithm in [1] to an MAR scenario for the Metaverse.…”

“…Comparison with our ICDCS paper [1]. We extend the algorithm in [1] to an MAR scenario for the Metaverse. In [1], we do not consider any optimization variable or objective function related to MAR.…”

“…We extend the algorithm in [1] to an MAR scenario for the Metaverse. In [1], we do not consider any optimization variable or objective function related to MAR. The optimization problem in [1] is to minimize the total energy and time consumption.…”

“…In [1], we do not consider any optimization variable or objective function related to MAR. The optimization problem in [1] is to minimize the total energy and time consumption. In this work, we add one more optimization variable-video frame resolution in the context of MAR for the Metaverse.…”