Research map / Label-efficient and robust learning / Recognition and long-tail
Finding Representative Interpretations on Convolutional Neural Networks
Peter Cho-Ho Lam, Lingyang Chu, Maxim Torgonskiy, Jian Pei, Yong Zhang, Lanjun Wang
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 RI instead of Grad-CAM for reusing interpretations | VGG-19 interpretation · GC unseen images | mAD: 23.01 -> 12.70 | Improved |
| use RI instead of Grad-CAM | VGG-19 interpretation · RO unseen images | mAD: 35.34 -> 12.57 | Improved |
| use RI instead of Grad-CAM | VGG-19 interpretation · ASIRRA unseen images | mAD: 2.30 -> 2.00 | Improved |
| use RI instead of Grad-CAM | VGG-19 interpretation · FOOD unseen images | mAD: 9.22 -> 8.08 | Improved |
| use RI instead of Grad-CAM | VGG-19 interpretation · ASIRRA unseen images | mAI: 32.90 -> 30.72 | Hurt |
| use RI instead of Grad-CAM | VGG-19 interpretation · RO unseen images | mAI: 22.02 -> 27.54 | Improved |
| use RI instead of ACE | VGG-19 interpretation · RO unseen images | mAD: 65.36 -> 12.57 | Improved |
| use RI instead of Score-CAM | VGG-19 interpretation · FOOD unseen images | mAI: 19.56 -> 26.66 | Improved |
| increase number of sampled linear boundaries |Q| from small to 50 (ablation) | RI · ASIRRA, GC, RO, FOOD | ACU-GT / ACU-MD | Improved |
| increase |Q| beyond 50 (ablation) | RI · ASIRRA, GC, RO, FOOD | ACU-GT / ACU-MD | No clear effect |
| show RI interpretation to non-expert subjects (Test B vs A) | RI · Retina OCT A/B test | difference score | Improved |
Closest papers by meaning
- Network Dissection: Quantifying Interpretability of Deep Visual Representations (CVPR 2017 · similarity 0.89)
- Interpreting Deep Visual Representations via Network Dissection (TPAMI 2018 · similarity 0.88)
- PICNN: A Pathway towards Interpretable Convolutional Neural Networks (AAAI 2024 · similarity 0.88)
- Explaining Deep Convolutional Neural Networks via Latent Visual-Semantic Filter Attention (CVPR 2022 · similarity 0.88)
- Comparing the Decision-Making Mechanisms by Transformers and CNNs via Explanation Methods (CVPR 2024 · similarity 0.87)
- Representational Similarity via Interpretable Visual Concepts (ICLR 2025 · 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.
