← Search

David Schneider-Joseph

1 accepted papers

2023

LEACE: Perfect linear concept erasure in closed form

NeurIPS 2023poster

Concept erasure aims to remove specified features from a representation. It can improve fairness (e.g. preventing a classifier from using gender or race) and interpretability (e.g. removing a concept to observe changes in model behavior). We introduce LEAst-squares Concept Erasure (LEACE), a closed-…