Energy-Efficient Quantum Federated Learning Over Low-Altitude UAV Networks With Theoretical Guarantees

Le résumé fourni par la source

This paper studies a novel energy-efficient quantum federated learning (QFL) framework designed for low-altitude quantum-enabled unmanned aerial vehicle (UAV) networks. We develop a novel energy minimization problem by jointly considering UAV’s quantum computing power, computational frequency, transmit power, number of quantum noise measurements, and trajectory. The energy minimization problem we formulate is inherently non-convex and cannot be efficiently solved using conventional convex optimization methods in its original form. To overcome this challenge, we develop an iterative optimization approach that integrates block coordinate descent (BCD) with successive convex approximation (SCA) techniques to effectively find optimal solutions for QFL’s energy optimization. Moreover, we conduct a holistic convergence analysis of federated quantum model training, which reveals the important trade-off between the convergence rate and number of quantum noise measurements. Evaluations under different data settings demonstrate that our theoretical results are consistent with QFL model training simulations, thereby facilitating future QFL algorithm design. Extensive simulation results also show that our optimization methodology significantly improves energy efficiency over baseline methods.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.