Tail-Latency-Aware Federated Learning with Pinching Antenna: Latency, Participation, and Placement

2026-01-26

AI Summary

This work addresses the dual challenges of clock synchronization bottlenecks caused by tail latency and slow convergence due to rapid sampling under non-IID data in synchronous wireless federated learning. To minimize the expected time to reach a target accuracy, the authors jointly optimize the placement of tunable pinching antennas (PAs) and client participation strategies. By modeling the maximum round latency via order statistics and integrating a heterogeneity-aware convergence factor with Lagrangian duality, they uncover a tail-latency premium embedded in KKT recursions. This leads to the derivation of an intra-class square-root sampling law, two phase-transition conditions, and—novel to this study—a piecewise-envelope derivative characterization of PA placement along with an exact candidate enumeration algorithm. Simulations demonstrate that the proposed approach significantly enhances eligible client participation and achieves higher accuracy under realistic clock constraints.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Systems and Infrastructure for Web, Mobile and WoT: Federated Web and WoT systems, including distributed, federated and edge-based data processing
• Security and Privacy: Large-scale security measurements

Abstract

Straggler synchronization is a dominant wall-clock bottleneck in synchronous wireless federated learning (FL). Under non-IID data, however, aggressively sampling only fast clients may significantly slow convergence due to statistical heterogeneity. This paper studies PASS-enabled FL, where a radiating pinching antenna (PA) can be activated at an arbitrary position along a dielectric waveguide to reshape uplink latencies. We consider a joint optimization of PA placement and client participation to minimize the expected time-to-accuracy, coupling the exact expected maximum round latency via order statistics with a heterogeneity-aware convergence factor. We derive first-order optimality conditions that reveal an explicit tail-latency premium in the KKT recursion, quantifying how latency gaps are amplified by maximum-order-statistic synchronization. Under a latency-class structure, we obtain a within-class square-root sampling law and establish a two-class phase transition where slow-class participation collapses under an explicit heterogeneity-threshold condition as the per-round sample size grows. For PA placement, we prove a piecewise envelope-derivative characterization and provide an exact breakpoint-and-root candidate-enumeration procedure. Simulation results verify the theoretical findings and show that PASS enables more eligible participation, yielding higher wall-clock accuracy.

Problem

• straggler synchronization
• tail latency
• federated learning
• statistical heterogeneity
• non-IID data

Innovation

• Pinching Antenna
• Tail Latency
• Federated Learning
• Order Statistics
• Heterogeneity-Aware Optimization