Proxy
noun
say it: “PROK-see” /ˈprɑksi/
Something that stands in for something else; in machine learning, a feature that carries the information of another, such as a sensitive one the model was not given.
What is a proxy in a machine learning model?
A proxy is a feature that stands in for another, such as a ZIP code that tells neighborhoods, and so often race or income, apart. Leaving a sensitive feature out of a model doesn't help if a proxy for it stays in: the model learns the old gap through the proxy.
In a sentence
- “Dropping the district column didn't help, because a proxy for it stayed in.”
- “Work hours turned out to be a proxy for district.”
Level: Journeyman · More AI Basics terms at this level
Updated