Research map / Learning theory and algorithms / Prediction and decision losses
A Holistic Approach to Unifying Automatic Concept Extraction and Concept Importance Estimation
Thomas FEL, Victor Boutin, Louis Béthune, Remi Cadene, Mazda Moayeri, Léo Andéol, Mathieu Chalvidal, Thomas Serre
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use PCA rather than KMeans for concept extraction | Concept extraction on the penultimate layer · ImageNet, 10 classes | Relative ℓ2: 0.60 / 0.54 / 0.73 vs 0.72 / 0.66 / 0.84 | Improved |
| Use KMeans rather than PCA for concept extraction | Concept extraction on the penultimate layer · ImageNet, 10 classes | Sparsity: 0.95 / 0.95 / 0.95 vs 0.00 / 0.00 / 0.0 | Improved |
| Use KMeans rather than PCA for concept extraction | Concept extraction on the penultimate layer · ImageNet, 10 classes | Stability: 0.07 / 0.08 / 0.04 vs 0.41 / 0.38 / 0.43 | Improved |
| Use NMF rather than PCA for concept extraction | Concept extraction on the penultimate layer · ImageNet, 10 classes | FID: 0.38 / 0.21 / 0.24 vs 0.47 / 0.17 / 0.24 | Improved |
| Use NMF rather than KMeans for concept extraction | Concept extraction on the penultimate layer · ImageNet, 10 classes | OOD: 1.98 / 0.29 / 0.15 vs 1.76 / 0.29 / 0.15 | Mixed |
| Decompose concepts in the last layer rather than earlier blocks | Concept extraction and concept importance estimation · ImageNet, ResNet50 blocks B2–B5 | C-Deletion, C-Insertion, and C-μFidelity | Improved |
| Use Gradient Input, Integrated Gradients, Occlusion, or RISE for concept importance in the penultimate layer | Concept attribution · Penultimate-layer decomposition | C-Deletion, C-Insertion, and C-μFidelity | Improved |
Closest papers by meaning
- ICE: Intrinsic Concept Extraction from a Single Image via Diffusion Models (CVPR 2025 · similarity 0.87)
- Restyling Unsupervised Concept Based Interpretable Networks with Generative Models (ICLR 2025 · similarity 0.87)
- ConceptBed: Evaluating Concept Learning Abilities of Text-to-Image Diffusion Models (AAAI 2024 · similarity 0.86)
- FaCT: Faithful Concept Traces for Explaining Neural Network Decisions (NeurIPS 2025 · similarity 0.86)
- Testing Semantic Importance via Betting (NeurIPS 2024 · similarity 0.86)
- Probabilistic Conceptual Explainers: Trustworthy Conceptual Explanations for Vision Foundation Models (ICML 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
