Skip to content
/ FedCD Public

A repo with code for the final project of Harvard's CS242 (Computing at Scale) on a new Federated Learning scheme.

Notifications You must be signed in to change notification settings

kavyakvk/FedCD

Repository files navigation

FedCD

Authors: Kavya Kopparapu, Eric Lin, Jessica Zhao

Federated learning has been widely applied to enable decentralized devices, which each have their own local data, to learn a shared model. However, learning from real-world data can be challenging, as it is rarely identically and independently distributed (IID) across edge devices (a key assumption for current high-performing and low-bandwidth algorithms). We present a novel approach, FedCD, which clones and deletes models to dynamically group devices with similar data. Experiments on the CIFAR-10 dataset show that FedCD achieves higher accuracy and faster convergence compared to a FedAvg baseline on non-IID data while incurring minimal computation, communication, and storage overheads.

Final Project for Harvard CS 242: Computing at Scale

Paper Accepted and Oral Presentation at the AI of Things Workshop at the ACM International Conference on Knowledge Discovery and Data Mining (San Diego, CA), August 2020: https://aiotworkshop.github.io/accepted-papers.html

Additional Information: Paper on Arxiv https://arxiv.org/abs/2006.09637

About

A repo with code for the final project of Harvard's CS242 (Computing at Scale) on a new Federated Learning scheme.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors