Research map / Label-efficient and robust learning / Recognition and long-tail
Explaining Neural Networks Semantically and Quantitatively
Runjin Chen, Hao Chen, Jie Ren, Ge Huang, Quanshi Zhang
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 |
|---|---|---|---|
| add prior-weight losses to the distillation-only explainer | explainer, AlexNet performer · Pascal-Part animal classification | entropy of contributions: 3.840 -> 5.136 | Improved |
| add prior-weight losses to the distillation-only explainer | explainer, VGG-S performer · Pascal-Part animal classification | entropy of contributions: 4.078 -> 5.831 | Improved |
| add prior-weight losses to the distillation-only explainer | explainer, VGG-16 performer · Pascal-Part animal classification | entropy of contributions: 4.392 -> 5.932 | Improved |
| add prior-weight losses to the distillation-only explainer | explainer, VGG-16 face attribute CNN · CelebA face attributes | entropy of contributions: 3.142 -> 3.201 | Improved |
| add prior-weight losses to the distillation-only explainer | explainer · Pascal-Part object parts | error of estimated part contribution: 0.164 -> 0.089 | Improved |
| add prior-weight losses to the distillation-only explainer | explainer · Pascal-Part eye part | error of estimated part contribution: 0.399 -> 0.181 | Improved |
| replace the performer by the explainer for prediction | VGG-16 face attribute CNN · CelebA face attributes | classification accuracy: 90.2 -> 85.9 | Hurt |
| replace the performer by the explainer for prediction | AlexNet · Pascal-Part animal classification | classification accuracy: 93.9 -> 92.6 | Hurt |
| replace the performer by the explainer for prediction | VGG-16 · Pascal-Part animal classification | classification accuracy: 95.4 -> 95.6 | No clear effect |
| use more visual concepts for explanation (ablation) | explainer · CelebA face attributes | explanation capacity | Improved |
Closest papers by meaning
- Interpretable Convolutional Neural Networks (CVPR 2018 · similarity 0.89)
- Explainable Neural Networks with Guarantee: A Sparse Estimation Approach (AAAI 2025 · similarity 0.88)
- Explainability in Graph Neural Networks: A Taxonomic Survey (TPAMI 2022 · similarity 0.88)
- Q-SENN: Quantized Self-Explaining Neural Networks (AAAI 2024 · similarity 0.88)
- Automated Natural Language Explanation of Deep Visual Neurons with Large Models (Student Abstract) (AAAI 2024 · similarity 0.88)
- Explaining, Fast and Slow: Abstraction and Refinement of Provable Explanations (ICML 2025 · similarity 0.87)
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.
