Research map / Architectures and efficiency / RNNs and normalization
Fast Finite Width Neural Tangent Kernel
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 explicit Jacobian contraction with structured derivatives | NTK computation for FCNs · FCN | time complexity: Jacobian contraction: $\n^2\o^2\t\w^2 + \n^2\o^3\w$; structured derivatives: $\n\o\t\w^2 + \n^2\o^2\t\w + \n^2\o^3$ | Helped |
| Replace explicit Jacobian contraction with structured derivatives | NTK computation for FCNs · FCN | memory complexity: $\n^2\o^2+\n\o\w^2+\n\o^2\w + \n\t\w+\t\w^2$ -> $\n^2\o^2+\n\o\w+\n\t\w+\t\w^2$ | Helped |
| Replace explicit Jacobian contraction with NTK-vector products | NTK computation for FCNs · FCN | time complexity: $\n^2\o^2\t\w^2 + \n^2\o^3\w$ -> $\n^2\o\t\w^2 + \n^2\o^2\w$ | Helped |
| Replace explicit Jacobian contraction with NTK-vector products | NTK computation for CNNs · CNN | time complexity | Mixed |
| Use structured derivatives instead of Jacobian contraction | NTK computation for CNNs · CNN | time complexity | Helped |
| Use structured derivatives instead of NTK-vector products | NTK computation · ResNets and other ImageNet models | wall-clock time | Helped |
Closest papers by meaning
- Neural Tangent Kernel Beyond the Infinite-Width Limit: Effects of Depth and Initialization (ICML 2022 · similarity 0.93)
- Finite-Width Neural Tangent Kernels from Feynman Diagrams (ICML 2026 · similarity 0.92)
- Scaling Neural Tangent Kernels via Sketching and Random Features (NeurIPS 2021 · similarity 0.92)
- Optimized Random Features for the Neural Tangent Kernel (Student Abstract) (AAAI 2025 · similarity 0.91)
- The Surprising Effectiveness of Infinite-Width NTKs for Characterizing and Improving Model Training (AAAI 2025 · similarity 0.90)
- On the Impacts of the Random Initialization in the Neural Tangent Kernel Theory (NeurIPS 2024 · similarity 0.89)
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.