US 12,147,500 B2
Privacy-sensitive training of user interaction prediction models
Lukas Zilka, Zurich (CH)
Assigned to GOOGLE LLC, Mountain View, CA (US)
Filed by GOOGLE LLC, Mountain View, CA (US)
Filed on Jul. 12, 2023, as Appl. No. 18/350,860.
Application 18/350,860 is a continuation of application No. 17/939,492, filed on Sep. 7, 2022, granted, now 11,741,191.
Application 17/939,492 is a continuation of application No. 16/393,777, filed on Apr. 24, 2019.
Prior Publication US 2023/0350978 A1, Nov. 2, 2023
This patent is subject to a terminal disclaimer.
Int. Cl. G06F 18/214 (2023.01); G06F 17/18 (2006.01); G06F 18/2415 (2023.01); G06N 3/02 (2006.01); G06N 20/00 (2019.01); G06V 10/46 (2022.01)
CPC G06F 18/2148 (2023.01) [G06F 17/18 (2013.01); G06F 18/2415 (2023.01); G06N 3/02 (2013.01); G06N 20/00 (2019.01); G06V 10/46 (2022.01); G06V 10/473 (2022.01)] 16 Claims
OG exemplary drawing
 
1. A method comprising:
generating training examples for training a local instance of a machine learning model based on interaction of a user of a user device with data elements presented to the user on webpages identified by search results responsive to search queries submitted by the user by:
determining that the user of the user device: submits a particular search query to a search system, receives a set of search results responsive to the particular search query from the search system, navigates to a webpage identified by a particular search result, and interacts with a particular data element presented on the webpage, and
generating a training example that includes a training input, and a target output that should be generated by the machine learning model by processing the training input, wherein the training input comprises the particular search query, the webpage, and the particular data element, and wherein the target output comprises data indicating that the user interacted with the particular data element;
training the local instance of the machine learning model on the training examples, using machine learning training techniques, to determine an update to current parameter values of the machine learning model received from a global training system; and
transmitting, to the global training system, parameter update data defining the update to the current parameter values of the machine learning model,
wherein the global training system is configured to:
receive parameter update data from a plurality of user devices, the user device being one of the plurality of user devices,
update the current parameter values of the machine learning model using the parameter update data received from the plurality of user devices, and
use the machine learning model in ranking search results.