Invertible Concept-based Explanations for CNN Models with Non-negative Concept Activation Vectors
This work addresses the explainability challenge for CNN models in computer vision, offering a method to improve interpretability for users in critical domains, though it is incremental as it builds on prior concept activation vector approaches.
The authors tackled the problem of explaining CNN models by proposing an invertible concept-based explanation framework that uses non-negative concept activation vectors, achieving superior interpretability and fidelity in computational and human subject experiments.
Convolutional neural network (CNN) models for computer vision are powerful but lack explainability in their most basic form. This deficiency remains a key challenge when applying CNNs in important domains. Recent work on explanations through feature importance of approximate linear models has moved from input-level features (pixels or segments) to features from mid-layer feature maps in the form of concept activation vectors (CAVs). CAVs contain concept-level information and could be learned via clustering. In this work, we rethink the ACE algorithm of Ghorbani et~al., proposing an alternative invertible concept-based explanation (ICE) framework to overcome its shortcomings. Based on the requirements of fidelity (approximate models to target models) and interpretability (being meaningful to people), we design measurements and evaluate a range of matrix factorization methods with our framework. We find that non-negative concept activation vectors (NCAVs) from non-negative matrix factorization provide superior performance in interpretability and fidelity based on computational and human subject experiments. Our framework provides both local and global concept-level explanations for pre-trained CNN models.