US Patent Application 17735020. DECENTRALIZED CROSS-NODE LEARNING FOR AUDIENCE PROPENSITY PREDICTION simplified abstract

DECENTRALIZED CROSS-NODE LEARNING FOR AUDIENCE PROPENSITY PREDICTION

Organization Name

MICROSOFT TECHNOLOGY LICENSING, LLC

Inventor(s)

Boyi Chen of Cupertino CA (US)

Tong Zhou of Sunnyvale CA (US)

Siyao Sun of Santa Clara CA (US)

Lijun Peng of Mountain View CA (US)

Xinruo Jing of Foster City CA (US)

Vakwadi Thejaswini Holla of San Jose CA (US)

Yi Wu of Palo Alto CA (US)

Pankhuri Goyal of San Jose CA (US)

Souvik Ghosh of Saratoga CA (US)

Zheng Li of Palo Alto CA (US)

Yi Zhang of Los Altos CA (US)

Onkar A. Dalal of Santa Clara CA (US)

Jing Wang of Los Altos CA (US)

Aarthi Jayaram of Palo Alto CA (US)

DECENTRALIZED CROSS-NODE LEARNING FOR AUDIENCE PROPENSITY PREDICTION - A simplified explanation of the abstract

This abstract first appeared for US patent application 17735020 titled 'DECENTRALIZED CROSS-NODE LEARNING FOR AUDIENCE PROPENSITY PREDICTION

Simplified Explanation

The disclosed technologies involve a process for combining data from different sources to improve a trained model.

• The process starts by receiving a trained model and data from a first-party system.
• A third-party data set is also received and combined with the first-party data set in a protected environment called a data clean room.
• The combined data set is then used to fine-tune the trained model, resulting in a third-party tuned model.
• The model parameter data learned during the tuning process is sent to an aggregator node.
• The aggregator node provides a globally tuned version of the trained model.
• This globally tuned model is applied to a second third-party data set, producing a scored third-party data set.
• Finally, the scored third-party data set is provided to a content distribution service of the first-party system.

Original Abstract Submitted

Embodiments of the disclosed technologies receive a first-party trained model and a first-party data set from a first-party system into a protected environment, receive a first third-party data set into the protected environment, and, in a data clean room, joining the first-party data set and the first third-party data set to create a joint data set for the particular segment, tuning a first-party trained model with the joint data set to create a third-party tuned model, sending model parameter data learned in the data clean room as a result of the tuning to an aggregator node, receiving a globally tuned version of the first-party trained model from the aggregator node, applying the globally tuned version of the first-party trained model to a second third-party data set to produce a scored third-party data set, and providing the scored third-party data set to a content distribution service of the first-party system.