首页|Patent Application Titled 'Counterfactual Policy Evaluation of Model Performance ' Published Online (USPTO 20240220859)

Patent Application Titled 'Counterfactual Policy Evaluation of Model Performance ' Published Online (USPTO 20240220859)

扫码查看
Reporters obtained the following quote from the background information supplied by the inventors:“Online systems use content selection frameworks to select con tent to present to their users. A contentselection framework is a set of rules, policies, and models that are used to select which content to presentto which users. For example, a content selection framework generally includes a machine-l earning modelthat has been trained to score content items, eligibility criteria that dictate which users are eligible toreceive certain content items, or hype rparameters that specify when and how content items are presentedto users. Onli ne systems may test a set of candidate content selection frameworks to determine whichcandidate selection framework performs best for users. However, testing a candidate selection frameworkcan require significant time or resource investme nts by the online system, which means that testing newcandidate selection frame works tends to be relatively rare.”

CyborgsEmerging TechnologiesMachine LearningPatent Application

2024

Robotics & Machine Learning Daily News

Robotics & Machine Learning Daily News

ISSN:
年,卷(期):2024.(Jul.22)