Advancements in Federated Learning and Wireless Communication Integration

The recent developments in the field of federated learning (FL) and wireless communication technologies highlight a significant shift towards addressing fairness, efficiency, and robustness in distributed learning environments. Innovations are particularly focused on overcoming challenges related to data heterogeneity, communication constraints, and the integration of FL with emerging wireless technologies. A notable trend is the exploration of over-the-air computation and federated meta-learning to enhance model fairness and customization for diverse tasks. Additionally, the integration of FL with advanced wireless communication techniques, such as non-orthogonal multiple access (NOMA) and rate-splitting multiple access (RSMA), is gaining traction for improving network efficiency and security. The application of game theory and reinforcement learning in optimizing network operations and FL processes also represents a key area of advancement, aiming to achieve equilibrium and efficient resource allocation in complex, dynamic environments.

Noteworthy Papers

  • Over-the-Air Fair Federated Learning via Multi-Objective Optimization: Introduces a novel algorithm leveraging over-the-air computation to achieve fairness in FL, demonstrating superior performance in fairness and robustness.
  • Optimizing Value of Learning in Task-Oriented Federated Meta-Learning Systems: Proposes a task-oriented FML framework with a novel metric for assessing individual training needs, significantly outperforming baseline schemes.
  • IEEE 802.11bn Multi-AP Coordinated Spatial Reuse with Hierarchical Multi-Armed Bandits: Explores the use of reinforcement learning for optimizing C-SR group selection in Wi-Fi 8, identifying UCB as particularly effective.
  • Constrained Over-the-Air Model Updating for Wireless Online Federated Learning with Delayed Information: Presents COMUDO, a method for efficient model updating in wireless online FL with delayed information, showing substantial accuracy gains.
  • Over-the-Air FEEL with Integrated Sensing: Joint Scheduling and Beamforming Design: Develops a robust design for OTA-FEEL that leverages sensing capabilities to mitigate interference, ensuring quality model aggregation.
  • A Federated Deep Learning Framework for Cell-Free RSMA Networks: Proposes a novel cell-free network architecture integrating RSMA and FL, demonstrating high performance with enhanced security.
  • Smooth Handovers via Smoothed Online Learning: Introduces a SOL-based algorithm for optimizing handovers in cellular networks, providing robust dynamic regret guarantees.
  • Multiplayer Federated Learning: Reaching Equilibrium with Less Communication: Introduces MpFL, modeling FL clients as game-theoretic players, and proposes PEARL-SGD for reaching equilibrium with reduced communication.

Sources

Over-the-Air Fair Federated Learning via Multi-Objective Optimization

Optimizing Value of Learning in Task-Oriented Federated Meta-Learning Systems

IEEE 802.11bn Multi-AP Coordinated Spatial Reuse with Hierarchical Multi-Armed Bandits

The Bakers and Millers Game with Restricted Locations

Constrained Over-the-Air Model Updating for Wireless Online Federated Learning with Delayed Information

Over-the-Air FEEL with Integrated Sensing: Joint Scheduling and Beamforming Design

A Federated Deep Learning Framework for Cell-Free RSMA Networks

On Stability and Learning of Competitive Equilibrium in Generalized Fisher Market Models: A Variational Inequality Approach

Smooth Handovers via Smoothed Online Learning

Multiplayer Federated Learning: Reaching Equilibrium with Less Communication

Built with on top of