Straggler-Resilient Federated Learning over A Hybrid Conventional and Pinching Antenna Network

Abstract--Leveraging pinching antennas in wireless network enabled federated learning (FL) can effectively mitigate t he common "straggler" issue in FL by dynamically establishing strong line-of-sight (LoS) links on demand. This letter pro poses a hybrid conventional and pinching antenna network (HCPAN) to significantly improve communication efficiency in the non - orthogonal multiple access (NOMA)-enabled FL system. With in this framework, a fuzzy logic-based client classification s cheme is first proposed to effectively balance clients' data contr ibutions and communication conditions. Given this classification, w e formulate a total time minimization problem to jointly opti mize pinching antenna placement and resource allocation. Due to the complexity of variable coupling and non-convexity, a de ep reinforcement learning (DRL)-based algorithm is develope d to effectively address this problem. Simulation results vali date the superiority of the proposed scheme in enhancing FL performa nce via the optimized deployment of pinching antenna.