US Patent Application 18218319. MIXED CLIENT-SERVER FEDERATED LEARNING OF MACHINE LEARNING MODEL(S) simplified abstract

MIXED CLIENT-SERVER FEDERATED LEARNING OF MACHINE LEARNING MODEL(S)

Organization Name

Google LLC

Inventor(s)

Françoise Beaufays of Mountain View CA (US)

Andrew Hard of Menlo Park CA (US)

Swaroop Indra Ramaswamy of Belmont CA (US)

Om Dipakbhai Thakkar of San Jose CA (US)

Rajiv Mathews of Sunnyvale CA (US)

MIXED CLIENT-SERVER FEDERATED LEARNING OF MACHINE LEARNING MODEL(S) - A simplified explanation of the abstract

This abstract first appeared for US patent application 18218319 titled 'MIXED CLIENT-SERVER FEDERATED LEARNING OF MACHINE LEARNING MODEL(S)

Simplified Explanation

This patent application is about a method for federated learning of machine learning models using gradients generated on client devices and a remote system.

• The method involves client devices processing local data using on-device ML models to generate predicted outputs and gradients.
• The client devices then transmit these gradients to a remote system.
• The remote system processes remote data from databases using global ML models to generate additional predicted outputs and gradients.
• The client gradients and remote gradients are used to update the global ML models or their weights.
• The updated global ML models or weights are then sent back to the client devices.

Original Abstract Submitted

Implementations disclosed herein are directed to federated learning of machine learning (“ML”) model(s) based on gradient(s) generated at corresponding client devices and a remote system. Processor(s) of the corresponding client devices can process client data generated locally at the corresponding client devices using corresponding on-device ML model(s) to generate corresponding predicted outputs, generate corresponding client gradients based on the corresponding predicted outputs, and transmit the corresponding client gradients to the remote system. Processor(s) of the remote system can process remote data obtained from remote database(s) using global ML model(s) to generate additional corresponding predicted outputs, generate corresponding remote gradients based on the additional corresponding predicted outputs. Further, the remote system can utilize the corresponding client gradients and the corresponding remote gradients to update the global ML model(s) or weights thereof. The updated global ML model(s) and/or the updated weights thereof can be transmitted back to the corresponding client devices.