scientific-federated-learning

What problem does it solve?

Federated learning across multiple organizations to train models without sharing raw data, enabling privacy-preserving collaborative ML.

Core Features & Use Cases

• Flower-based FL pipeline with FedAvg/FedProx/FedOpt aggregation strategies
• Differential privacy via DP-SGD and support for non-IID data splits
• Efficient cross-site coordination and workflow orchestration for scientific data analysis
• Use Case: Research institutions and hospitals collaboratively train models without exposing sensitive data.

Quick Start

Launch a Flower-based federated learning pipeline across clients and apply DP-SGD with non-IID data splits to start training.