Federated Learning: Theory and Practice

Project Overview

Advisor: Prof. Avi Mohan

Conducting research on federated learning convergence theory through both theoretical analysis and empirical validation using object detection tasks.

Research Focus

This work investigates how different system configurations affect convergence in federated learning settings with heterogeneous data distributions.

Key Areas:

• Convergence analysis in distributed learning
• Heterogeneous data handling
• Practical federated learning implementations

Methodology

Dual-track research approach combining:

• Empirical validation on computer vision tasks
• Controlled experiments with varying data heterogeneity levels