View source for US Patent Application 17848947. SYSTEM(S) AND METHOD(S) FOR JOINTLY LEARNING MACHINE LEARNING MODEL(S) BASED ON SERVER DATA AND CLIENT DATA simplified abstract

=SYSTEM(S) AND METHOD(S) FOR JOINTLY LEARNING MACHINE LEARNING MODEL(S) BASED ON SERVER DATA AND CLIENT DATA=


==Organization Name==

[[Google LLC]]


==Inventor(s)==

[[Sean Augenstein of San Mateo CA (US)]]

[[Andrew Hard of Menlo Park CA (US)]]

[[Kurt Partridge of San Francisco CA (US)]]

[[Rajiv Mathews of Sunnyvale CA (US)]]

[[Lin Ning of San Jose CA (US)]]

[[Karan Singhal of Roslyn NY (US)]]

==SYSTEM(S) AND METHOD(S) FOR JOINTLY LEARNING MACHINE LEARNING MODEL(S) BASED ON SERVER DATA AND CLIENT DATA - A simplified explanation of the abstract==


This abstract first appeared for US patent application 17848947 titled 'SYSTEM(S) AND METHOD(S) FOR JOINTLY LEARNING MACHINE LEARNING MODEL(S) BASED ON SERVER DATA AND CLIENT DATA

==Simplified Explanation==

The patent application is about techniques to prevent catastrophic forgetting in federated learning of global machine learning models.
* The implementation identifies a global machine learning model that is initially trained on a remote server using server data.
* The server-based data includes EWC loss terms, client augmenting gradients, server augmenting gradients, and server-based data.
* The global ML model and server-based data are transmitted to multiple client devices.
* The client devices generate client gradients based on processing predicted output using the global ML model and the server-based data.
* The client gradients are then transmitted back to the remote server.
* An updated global ML model is generated based on the client gradients.


==Original Abstract Submitted==

Implementations disclosed herein are directed to various techniques for mitigating and/or preventing catastrophic forgetting in federated learning of global machine learning (ML) models. Implementations may identify a global ML model that is initially trained at a remote server based on a server data set, determine server-based data for global weight(s) of the global ML model, and transmit the global ML model and the server-based data to a plurality of client devices. The server-based data may include, for example, EWC loss term(s), client augmenting gradients, server augmenting gradients, and/or server-based data. Further, the plurality client devices may generate, based on processing corresponding predicted output and using the global ML model, and based on the server-based data, a corresponding client gradient, and transmit the corresponding client gradient to the remote server. Implementations may further generate an updated global ML model based on at least the corresponding client gradients.