Google Federated Learning

New way to share data for training prediction models. 

Google's Federated Learning Architecture Can Enhance Privacy While Ending the Centralized Dataset     Analytics India     By Abhijeet Katte

Google engineers recently announced Federated Learning, a new cloud architecture for processing machine learning data for models that are trained from user interactions on mobile devices. In the past, machine learning has required that data be in the data center or on the machine on which the model is being trained. Federated Learning uses mobile phones to collaborate and learn a shared prediction model, with training data remaining on the device and not transmitted to the cloud. With the new architecture, the mobile device downloads the current model and improves it by learning from data on the phone. The phone then generates a summary of what it has learned, which is sent to the cloud and aggregated with other user summaries to refine the shared model. This type of architecture enables smarter models, reduced latency, and lower power consumption while ensuring data privacy through encryption. The researchers say Federated Learning "has only scratched the surface of what is possible" with this type of distributed prediction model. ... "