FedX-mod
Federated learning with SimSiam as a lightweight alternative to contrastive methods, evaluated on CIFAR-10 and SVHN.
Course project exploring federated learning under communication and computation constraints. The original FedX approach uses contrastive self-supervised learning across distributed clients; this work substitutes SimSiam — a self-supervised method that requires no negative pairs — as a lighter alternative.
Performance was benchmarked against the original contrastive method on CIFAR-10 and SVHN image classification datasets.