google-parfait/tensorflow-federated

An open-source framework for machine learning and other computations on decentralized data.

What it solves

It addresses the challenge of training machine learning models on decentralized data without requiring that sensitive data be uploaded to a central server. This allows for collaborative model training while keeping user data local and private.

How it works

TFF uses a federated learning approach where a shared global model is trained across many participating clients. It provides two primary layers of interfaces:

  • Federated Learning (FL) API: A high-level layer (tff.learning) that allows developers to apply existing federated training and evaluation implementations to TensorFlow models.
  • Federated Core (FC) API: A low-level, strongly-typed functional programming environment that combines TensorFlow with distributed communication operators to express novel federated algorithms.

Who it’s for

It is designed for researchers and developers who want to experiment with federated learning algorithms, implement non-learning computations like aggregated analytics over decentralized data, or deploy federated computations to various runtime environments.

Highlights

  • Supports both high-level training/evaluation APIs and low-level algorithm design tools.
  • Enables training of models (e.g., mobile keyboard prediction) without uploading sensitive data.
  • Includes a single-machine simulation runtime for rapid experimentation.
  • Declaratively expresses computations for deployment across diverse environments.

Related

  • Project
    tensorflow/tensorflow

    An end-to-end open source platform for machine learning that provides a flexible ecosystem of tools and libraries for building and deploying ML-powered applications.

  • Dispatch
    Creating Privacy Preserving AI with Substra

    Hugging Face and Substra highlight how federated learning enables the training of AI models across decentralized data sources to maintain privacy and security, particularly in sensitive domains like healthcare.

  • Project
    LongxingTan/Time-series-prediction

    A TensorFlow-based time series package that provides a unified interface for prediction, classification, and anomaly detection using state-of-the-art deep learning models.

  • Project
    NVIDIA/NVFlare

    An open-source Python SDK for federated learning that enables secure, privacy-preserving machine learning collaboration across distributed parties without sharing raw data.

  • Project
    apple/pfl-research

    A Python framework for simulating private federated learning experiments, supporting multiple ML frameworks and differential privacy mechanisms.