Federated Deep Equilibrium Learning on Edge Networks

This project studies Federated Deep Equilibrium Models for training across heterogeneous clients and datasets under constrained edge environments.

I implemented FeDEQ components, explored communication-efficient aggregation techniques, and evaluated performance across NLP and vision benchmarks under non-IID settings.

Highlights

• Implemented federated deep equilibrium learning experiments for heterogeneous clients.
• Designed communication-efficient aggregation techniques to improve convergence and reduce energy use.
• Evaluated models across NLP and vision benchmarks under non-IID data splits.
• Supported a manuscript submitted to the IEEE Internet of Things Journal (Le et al., 2025).