نوع مقاله : علمی-پژوهشی
نویسندگان
1 عضو هیات علمی دانشگاه صنعتی نوشیروانی بابل
2 گروه مخابرات، دانشکده مهندسی برق و کامپیوتر، دانشگاه صنعتی نوشیروانی بابل
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
Federated learning is an emerging paradigm in distributed machine learning, where users collaboratively train a global model by updating local models and transmitting gradients—without sharing their raw data. This learning approach is particularly valuable in privacy-sensitive and distributed environments. Despite its advantages, communication errors during wireless model transmission remain one of the key challenges in federated learning systems. In this paper, we study beamforming design and resource allocation in a MIMO-SWIPT network with users participating in federated learning. The main objective is to minimize the learning loss function while considering wireless physical-layer constraints. To this end, we establish a relationship between physical parameters of the wireless system and the learning loss function, from which we derive a new metric called the "gap function." A joint optimization problem is then formulated over transmission time, power allocation, and beamforming vector to minimize this gap function. An iterative algorithm based on convex optimization is proposed to determine the optimal beamforming vector and time allocation. The proposed algorithms are evaluated from two perspectives: the objective function of the formulated optimization problem and the performance of federated learning using a specific dataset. Simulation results demonstrate that the proposed approach can effectively improve learning performance in federated wireless networks.
کلیدواژهها [English]
)