Dynamic Adaptive Federated Learning for mmWave Sector Selection

2025-10-05

AI Summary

To address high latency and signaling overhead caused by conventional beam sector selection in millimeter-wave (mmWave) communications, this paper proposes a dynamic adaptive federated learning framework tailored for autonomous networks. The method introduces a layer-wise dynamic aggregation and clustering mechanism, integrating layer importance evaluation with cross-node collaborative training to enhance model generalization while mitigating overfitting. Crucially, it enables selective aggregation of critical model layers, significantly improving training efficiency and deployment adaptability. Experiments on a real-world multimodal dataset demonstrate that the proposed approach achieves a 6.76% improvement in model accuracy, an 84.04% reduction in inference latency, and a 52.20% compression in model parameter count—effectively balancing accuracy, latency, and resource consumption.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Multiagent Systems: Multiagent Learning
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Beamforming techniques use massive antenna arrays to formulate narrow Line-of-Sight signal sectors to address the increased signal attenuation in millimeter Wave (mmWave). However, traditional sector selection schemes involve extensive searches for the highest signal-strength sector, introducing extra latency and communication overhead. This paper introduces a dynamic layer-wise and clustering-based federated learning (FL) algorithm for beam sector selection in autonomous vehicle networks called enhanced Dynamic Adaptive FL (eDAFL). The algorithm detects and selects the most important layers of a machine learning model for aggregation in the FL process, significantly reducing network overhead and failure risks. eDAFL also considers intra-cluster and inter-cluster approaches to reduce overfitting and increase the abstraction level. We evaluate eDAFL on a real-world multi-modal dataset, demonstrating improved model accuracy by approximately 6.76% compared to existing methods, while reducing inference time by 84.04% and model size by up to 52.20%.

Problem

• Reduces latency in mmWave beam sector selection
• Minimizes communication overhead for autonomous vehicles
• Addresses overfitting in federated learning models

Innovation

• Uses dynamic layer-wise federated learning for beam selection
• Employs clustering approaches to reduce overfitting risks
• Selects important model layers to minimize network overhead