Skip to content

Self-Supervised Personalized FL

This category aims to achieve personalized federated learning by introducing self-supervised learning (SSL) to the training process. With SSL, an encoder model is trained to learn representations from unlabeled data. A higher performance can be achieved in subsequent tasks with the trained encoder. Only the encoder model is globally aggregated and shared during the regular training process. After reaching convergence, each client can download the trained global model to extract features from local samples.

Calibre

cd examples/ssl/calibre
uv run calibre.py -c ../configs/calibre_CIFAR10_resnet18.toml

Reference: S. Chen, N. Su, B. Li. "Calibre: Towards Fair and Accurate Personalized Federated Learning with Self-Supervised Learning," in Proc. IEEE ICDCS 2024.

SimCLR

cd examples/ssl/simclr
uv run simclr.py -c ../configs/simclr_CIFAR10_resnet18.toml

Reference: Chen et al., "A Simple Framework for Contrastive Learning of Visual Representations," in Proc. ICML, 2020.

BYOL

cd examples/ssl/byol
uv run byol.py -c ../configs/byol_CIFAR10_resnet18.toml

Reference: Grill et al., "Bootstrap Your Own Latent A New Approach to Self-Supervised Learning," in Proc. NeurIPS, 2020.

SimSiam

cd examples/ssl/simsiam
uv run simsiam.py -c ../configs/simsiam_CIFAR10_resnet18.toml

Reference: Chen et al., "Exploring Simple Siamese Representation Learning," in Proc. CVPR, 2021.

MoCoV2

cd examples/ssl/moco
uv run mocov2.py -c ../configs/mocov2_CIFAR10_resnet18.toml

Reference: Chen et al., "Improved Baselines with Momentum Contrastive Learning," in ArXiv, 2020.

SwAV

cd examples/ssl/swav
uv run swav.py -c ../configs/swav_CIFAR10_resnet18.toml

Reference: Caron et al., "Unsupervised Learning of Visual Features by Contrasting Cluster Assignments," in Proc. NeurIPS, 2022.

SMoG

cd examples/ssl/smog
uv run smog.py -c ../configs/smog_CIFAR10_resnet18.toml

Reference: Pang et al., "Unsupervised Visual Representation Learning by Synchronous Momentum Grouping," in Proc. ECCV, 2022.

FedEMA

cd examples/ssl/fedema
uv run fedema.py -c ../configs/fedema_CIFAR10_resnet18.toml

Reference: Zhuang et al., "Divergence-Aware Federated Self-Supervised Learning," in Proc. ICLR, 2022.