Robotics & Machine Learning Daily News2024,Issue(Aug.19) :97-100.

Patent Application Titled 'Outcome-Guided Counterfactuals from a Jointly Trained Generative Latent Space' Published Online (USPTO 20240256858)

Robotics & Machine Learning Daily News2024,Issue(Aug.19) :97-100.

Patent Application Titled 'Outcome-Guided Counterfactuals from a Jointly Trained Generative Latent Space' Published Online (USPTO 20240256858)

扫码查看

Abstract

Reporters obtained the following quote from the background information supplied by the inventors:“Counterfactual generation is the process of creating hypothet ical scenarios that differ from an observedevent in a specific way. The hypothe tical scenarios are often used to understand the causal relationshipsbetween di fferent factors and to explore what could have been. The traditional approach to counterfactualgeneration involves pulling examples from a repository of observ ed instances. The traditional approachmay be effective if the repository contai ns a large number of examples that are similar to the query.However, it may be difficult to find a counterfactual that is similar enough to the query, especial ly ifthe query is for a rare or unusual event. An alternative approach to count erfactual generation is to useincremental changes to scenes to generate a new c ounterfactual. The incremental changes approach maybe effective for generating counterfactuals for events that are not well-represented in the repository of observed instances.

Key words

Cyborgs/Emerging Technologies/Machine Learning/Patent Application

引用本文复制引用

出版年

2024
Robotics & Machine Learning Daily News

Robotics & Machine Learning Daily News

ISSN:
段落导航相关论文