Improving Interpretability of CNN Models Using Non-Negative Concept Activation Vectors

Abstract

Pre-print (2020)