FL-Development

This is a repo for implementing federated learning (FL) simulation studies in healthcare.

SOTA FL Algorithms by Year 📚 (updating)

2022

• FedLC -- Federated Learning with Label Distribution Skew via Logits Calibration (ICML'22)

• pFedLA -- Layer-Wised Model Aggregation for Personalized Federated Learning (CVPR'22)

• FedAP -- Personalized Federated Learning with Adaptive Batchnorm for Healthcare (IEEE'22)

• kNN-Per -- Personalized Federated Learning through Local Memorization (ICML'22)

• MetaFed -- MetaFed: Federated Learning among Federations with Cyclic Knowledge Distillation for Personalized Healthcare (IJCAI'22)

2021

• MOON -- Model-Contrastive Federated Learning (CVPR'21)

• FedBN -- FedBN: Federated Learning On Non-IID Features Via Local Batch Normalization (ICLR'21)

• FedGen -- Data-Free Knowledge Distillation for Heterogeneous Federated Learning (ICML'21)

• Ditto -- Ditto: Fair and Robust Federated Learning Through Personalization (ICML'21)

• pFedHN -- Personalized Federated Learning using Hypernetworks (ICML'21)

2020

• FedProx -- Federated Optimization in Heterogeneous Networks (MLSys'20)

• SCAFFOLD -- SCAFFOLD: Stochastic Controlled Averaging for Federated Learning (ICML'20)

• FedPer -- Federated Learning with Personalization Layers (AISTATS'20)

• Per-FedAvg -- Personalized Federated Learning with Theoretical Guarantees: A Model-Agnostic Meta-Learning Approach (NIPS'20)

2019

• FedMD -- FedMD: Heterogenous Federated Learning via Model Distillation (NIPS'19)

• FedAvgM -- Measuring the Effects of Non-Identical Data Distribution for Federated Visual Classification (ArXiv'19)

• CFL -- Clustered Federated Learning: Model-Agnostic Distributed Multi-Task Optimization under Privacy Constraints (ArXiv'19)

2017

• FedAvg -- Communication-Efficient Learning of Deep Networks from Decentralized Data (AISTATS'17)

Environment Requirements 🛠

pip install -r requirements.txt

Docker 🐳

Coming soon...

Dataset Partition 📂

This part aims to divide the entire datasets into Non-IID datasets according to the number of clients and the partition scheme (Dirichlet distribution).

# partition the MedMNIST dataset into 10 clients according to Dirichlet distribution with alpha=0.1
python generate_data.py -d medmnistS -a 0.1 -cn 10

Run FL Simulation 🏃‍♂️

# run FedAvg simulation on MedMNIST dataset
python fedavg.py -d medmnistS

• Note: ALL the methods are inherited from FedAvgServer and FedAvgClient.
• please check arguments_detail for more details.

Supported Tasks 📝

• Classification
• Reconstruction (coming soon...)
• Segmentation (coming soon...)

Supported Medical Datasets 🎨

• MedMNIST

• COVID-19

• NIH Chest X-ray 14

• FeTS 2022 (coming soon...)

To-Do List 📝

• support more medical datasets
• support more SOTA FL methods
• support more tasks
• docker support
• pre-trained weights
• ...