Research map / Learning theory and algorithms / Kernels and regression theory
Locality defeats the curse of dimensionality in convolutional teacher-student scenarios
Alessandro Favero, Francesco Cagnetta, Matthieu Wyart
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 convolutional (weight-shared) student instead of local student with same filter size | local kernel regression · Convolutional teacher, ridgeless, synthetic | learning curve exponent beta | No clear effect |
| student filter size s smaller than teacher filter size t | kernel regression student · Synthetic teacher-student | generalisation error | Hurt |
| student filter size s equal to input dimension d (no locality) | kernel regression student · Synthetic teacher-student | learning curve exponent | Hurt |
| use smaller student filter size s with s >= t | kernel regression student · Synthetic teacher-student | learning curve exponent beta | Improved |
| use smaller local filter size | local kernels · CIFAR-10 | learning curve decay | Improved |
Closest papers by meaning
- Role of Locality and Weight Sharing in Image-Based Tasks: A Sample Complexity Separation between CNNs, LCNs, and FCNs (ICLR 2024 · similarity 0.87)
- Learning from Few Samples: Transformation-Invariant SVMs with Composition and Locality at Multiple Scales (NeurIPS 2022 · similarity 0.86)
- Locality in Image Diffusion Models Emerges from Data Statistics (NeurIPS 2025 · similarity 0.86)
- Learning curves of generic features maps for realistic datasets with a teacher-student model (NeurIPS 2021 · similarity 0.86)
- Equivariant Neural Tangent Kernels (ICML 2025 · similarity 0.84)
- How Feature Learning Can Improve Neural Scaling Laws (ICLR 2025 · similarity 0.84)
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.
