Dynamic D2D-Assisted Federated Learning over O-RAN: Performance Analysis, MAC Scheduler, and Asymmetric User Selection

2024-04-09

arXiv.org

AI Summary

Existing federated learning (FL) research predominantly assumes static network conditions, overlooking the time-varying nature of wireless channel capacity and user data distributions. To address this gap in open radio access networks (O-RAN), we propose DCLM—a dynamic device-to-device (D2D) collaborative FL framework. DCLM is the first to systematically model multi-granularity system dynamics: (i) discrete-event-driven abrupt channel capacity changes ($mathscr{D}$-Events), and (ii) continuous dynamics of dataset evolution and model drift, captured via ordinary differential equations (ODEs) and partial differential inequalities (PDIs). The framework integrates hierarchical D2D training, O-RAN-native MAC scheduling, and an asymmetric user selection mechanism. We establish theoretical convergence guarantees and design an efficient non-convex optimization solver. Extensive simulations demonstrate that DCLM significantly improves model accuracy and resource efficiency under time-varying conditions, establishing a new network-aware FL paradigm.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization
• Search and Optimization: Distributed 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
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Existing studies on federated learning (FL) are mostly focused on system orchestration for static snapshots of the network and making static control decisions (e.g., spectrum allocation). However, real-world wireless networks are susceptible to temporal variations of wireless channel capacity and users' datasets. In this paper, we incorporate multi-granular system dynamics (MSDs) into FL, including (M1) dynamic wireless channel capacity, captured by a set of discrete-time events, called $mathscr{D}$-Events, and (M2) dynamic datasets of users. The latter is characterized by (M2-a) modeling the dynamics of user's dataset size via an ordinary differential equation and (M2-b) introducing dynamic model drift}, formulated via a partial differential inequality} drawing concrete analytical connections between the dynamics of users' datasets and FL accuracy. We then conduct FL orchestration under MSDs by introducing dynamic cooperative FL with dedicated MAC schedulers (DCLM), exploiting the unique features of open radio access network (O-RAN). DCLM proposes (i) a hierarchical device-to-device (D2D)-assisted model training, (ii) dynamic control decisions through dedicated O-RAN MAC schedulers, and (iii) asymmetric user selection. We provide extensive theoretical analysis to study the convergence of DCLM. We then optimize the degrees of freedom (e.g., user selection and spectrum allocation) in DCLM through a highly non-convex optimization problem. We develop a systematic approach to obtain the solution for this problem, opening the door to solving a broad variety of network-aware FL optimization problems. We show the efficiency of DCLM via numerical simulations and provide a series of future directions.

Problem

• Addressing dynamic wireless channel capacity in federated learning
• Modeling dynamic user datasets and their impact on accuracy
• Optimizing network-aware federated learning with non-convex constraints

Innovation

• Hierarchical D2D-assisted model training for FL
• Dynamic control via dedicated O-RAN MAC schedulers
• Asymmetric user selection for network-aware optimization