Research map / Architectures and efficiency / CNN design
Neural Similarity Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| replace standard convolution with dynamic diagonal neural similarity | CNN-9 · CIFAR-10 | error (%): 7.78 -> 6.85 | Helped |
| replace standard convolution with dynamic diagonal neural similarity | CNN-9 · CIFAR-100 | error (%): 28.95 -> 27.81 | Helped |
| replace standard convolution with dynamic diagonal neural similarity | CNN-10 · ImageNet-2012 val | Top-1 error (%): 42.72 -> 40.61 | Helped |
| replace standard convolution with dynamic diagonal neural similarity | CNN-10 · ImageNet-2012 val | Top-5 error (%): 19.11 -> 18.04 | Helped |
| use dynamic rather than static neural similarity | NSN · CIFAR-10 | error (%): 7.15 -> 6.85 | Helped |
| use dynamic rather than static neural similarity | NSN · CIFAR-100 | error (%): 28.35 -> 27.81 | Helped |
| use unconstrained rather than diagonal neural similarity (ablation) | static NSN · CIFAR-10 | error (%): 7.15 -> 7.38 | Hurt |
| use unconstrained rather than diagonal neural similarity (ablation) | static NSN · CIFAR-100 | error (%): 28.35 -> 28.11 | Helped |
| use unconstrained rather than diagonal neural similarity (ablation) | dynamic NSN · CIFAR-10 | error (%): 6.85 -> 6.5 | Helped |
| use unconstrained rather than diagonal neural similarity (ablation) | dynamic NSN · CIFAR-100 | error (%): 27.81 -> 28.02 | Hurt |
| use SphereNet instead of a CNN as the dynamic similarity predictor (ablation) | dynamic NSN · CIFAR-10 | error (%): 7.04 -> 6.85 | Helped |
| jointly learn kernel shape along with similarity (ablation) | static NSN · CIFAR-10 | error (%): 7.15 -> 6.92 | Helped |
Closest papers by meaning
- Decoupling Semantic Similarity from Spatial Alignment for Neural Networks. (NeurIPS 2024 · similarity 0.85)
- Tracing Representation Progression: Analyzing and Enhancing Layer-Wise Similarity (ICLR 2025 · similarity 0.84)
- Representational Similarity via Interpretable Visual Concepts (ICLR 2025 · similarity 0.84)
- Towards flexible perception with visual memory (ICML 2025 · similarity 0.84)
- Going Beyond Neural Network Feature Similarity: The Network Feature Complexity and Its Interpretation Using Category Theory (ICLR 2024 · similarity 0.83)
- ReSi: A Comprehensive Benchmark for Representational Similarity Measures (ICLR 2025 · similarity 0.83)
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.