Federated learning introduction and examples with TensorFlow Federated

Description

With the advent of connected devices with computation and storage capabilities, it’s now possible to run machine learning workflows entirely on device. Federated learning is an approach for training ML models across a fleet of participating devices without collecting the data in a central location. Federated learning improves upon the traditional, fully centralized approaches by reducing the costs and risks related to sensitive data handling, working better in bandwidth- and power-constrained environments, and providing a straightforward, effective mechanism for personalization at scale. It also puts users back in control of their data while still enabling developers to build intelligent applications that leverage insights from that data.

Federated learning is already used at scale by Google, where it powers predictive input capabilities in the Android keyboard (Gboard), on-device search for Pixel phones, and other experiences. The recent release of the open source TensorFlow Federated project enables all ML developers to experiment with federated technologies. Alex Ingerman offers an overview of federated learning, compares traditional and federated ML workflows, and explores the current and upcoming use cases for decentralized machine learning with examples from Google’s deployment of this technology. You’ll have the opportunity to get your hands dirty in a hands-on introduction to TensorFlow Federated as Alex shows you how it can be used to train federated models on decentralized datasets.