1# Federated Compute Platform 2 3This repository hosts code for executing federated programs and computations. 4 5## Definitions 6 7A *federated computation* is a set of processing steps that run on a server and 8set of clients, where each step is either 9 10* local processing of values on the client or server or 11* transport which moves values via 12 * broadcast (server-to-clients) 13 * select (client-server response) 14 * aggregate (clients-to-server). A federated computation invoked by a 15 central coordinator returns one or more aggregates. 16 17A *federated program* is a set of processing steps run by a central coordinator 18that include one or more invocations of federated computation, transformations 19of computation results, and releases of aggregate values to the engineer/analyst 20who invoked the program. 21 22To learn more about these concepts, check out: 23 24- [Federated learning comic book from Google AI](http://g.co/federated) 25- [Federated Learning: Collaborative Machine Learning without Centralized 26 Training 27 Data](https://ai.googleblog.com/2017/04/federated-learning-collaborative.html) 28- [Federated Analytics: Collaborative Data Science without Data Collection](https://ai.googleblog.com/2020/05/federated-analytics-collaborative-data.html) 29- [Towards Federated Learning at Scale: System Design](https://arxiv.org/abs/1902.01046) 30 (SysML 2019) 31- [Federated Program API in TFF](https://github.com/tensorflow/federated/blob/main/tensorflow_federated/python/program/README.md) 32 33## Infrastructure 34 35At Google, federated programs and computations are authored in 36[TensorFlow Federated](http://tensorflow.org/federated), compiled to deployable 37artifacts, and run in a distributed system consisting of a central coordinator, 38and a set of devices such as phones. The TFF repository contains infrastructure 39for authoring and simulating federated programs and computations. 40 41This repository hosts infrastructure for compiling and running federated 42programs and computations in the cross-device setting. We are actively working 43on open sourcing the core components of our production infrastructure, with a 44focus on privacy-sensitive code-paths such as the pipeline for compiling 45deployable artifacts from TFF computations, client-side processing, and 46server-side aggregation logic. 47 48As of 12/7/2022, parts of the repository - in particular, code in the `client/` 49directory, and the service & data format definitions in `proto/` - are used in 50production in Google's federated learning infrastructure. Other parts - notably, 51production server side infrastructure - have not yet been open sourced due to 52its dependencies on proprietary infrastructure, and we instead provide a 53reference / example server implementation in `demo/` for demonstration purposes. 54 55The best way to get started is to run the end-to-end demo 56`//fcp/demo:federated_program_test`, which will spin up example services, 57clients, and run a federated program; this test will cover the majority of the 58code in this repository. 59 60This is not an officially supported Google product. 61 62## Getting Started 63 64Please refer to the instructions in GETTING_STARTED.md. 65