Research map / Label-efficient and robust learning / OOD and anomaly detection
Building and Interpreting Deep Similarity Models
Oliver Eberle, Jochen Büttner, Florian Kräutli, Klaus‐Robert Müller, Matteo Valleriani, Grégoire Montavon
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 layer 17 instead of layer 31 features (ablation) | VGG-16 linear SVM · UFI | accuracy: 0.19 -> 0.62 | Improved |
| use layer 17 instead of layer 31 features (ablation) | VGG-16 linear SVM · LFW | accuracy: 0.75 -> 0.92 | Improved |
| use layer 17 instead of layer 31 features (ablation) | VGG-16 linear SVM · Sphaera | accuracy: 0.96 -> 0.98 | Improved |
| go from layer 5 to layer 31 in similarity model (ablation) | VGG-16 · UCF Sports Action | invariance: 2.30 -> 4.00 | Improved |
| use BiLRP instead of Saliency/Curvature baselines | similarity explanation · toy digit-matching | average cosine similarity to ground truth | Improved |
| use small nonzero gamma in BiLRP (vs gamma=0 Hessian x Product) (ablation) | BiLRP · toy digit-matching | average cosine similarity | Improved |
| replace VGG-16 layer 17 with engineered bigram network | table similarity · Sphaera tables | explanation quality | Improved |
Closest papers by meaning
- Why do These Match? Explaining the Behavior of Image Similarity Models (ECCV 2020 · similarity 0.83)
- Representational Similarity via Interpretable Visual Concepts (ICLR 2025 · similarity 0.82)
- Going Beyond Neural Network Feature Similarity: The Network Feature Complexity and Its Interpretation Using Category Theory (ICLR 2024 · similarity 0.82)
- When Does Closeness in Distribution Imply Representational Similarity? An Identifiability Perspective (NeurIPS 2025 · similarity 0.81)
- Decoupling Semantic Similarity from Spatial Alignment for Neural Networks. (NeurIPS 2024 · similarity 0.81)
- Interpretable Measures of Conceptual Similarity by Complexity-Constrained Descriptive Auto-Encoding (CVPR 2024 · similarity 0.81)
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.
