Avoid Forgetting by Preserving Global Knowledge Gradients in Federated Learning with Non-IID Data

Abhijit Chunduru, Majid Morafah, Mahdi Morafah, Vishnu Pandi Chellapandi, Ang Li

May, 2026

Abstract

Co-authored with collaborators from UCSD and Purdue, this work addresses catastrophic forgetting in federated learning through gradient-projection constraints.

Type

Journal article

Publication

Transactions on Machine Learning Research (under review)

Work on preserving global knowledge in non-IID federated learning through gradient-projection constraints.

Avoid Forgetting by Preserving Global Knowledge Gradients in Federated Learning with Non-IID Data

Abstract

Abhijit Chunduru

Master’s Student