Research map / Learning theory and algorithms / Sample complexity
Generalization Bounds for Neural Networks via Approximate Description Length
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 |
|---|---|---|---|
| Use approximate description length to bound generalization for networks with bounded weight magnitudes | Neural networks with bounded spectral and Frobenius norms · Bounded-weight neural networks | sample complexity: \tilde{O}\left( \frac{d^2R^2}{\epsilon^2} \right) -> \tilde{O}\left( \frac{dR^2}{\epsilon^2} \right) | Helped |
| Measure weight magnitude relative to fixed reference matrices instead of relative to zero | Neural networks with bounded weight changes from initialization · Neural networks around reference matrices | sample complexity: \tilde{O}\left( \frac{dR^2}{\epsilon^2} \right) | Helped |
| Use an approximate-description-length and shattering analysis to obtain an upper bound and a matching lower bound up to logarithmic factors | ReLU networks with bounded Frobenius norm · ReLU networks, depth two | fat-shattering dimension: \Omega\left(\frac{dR^2}{\log^2(d)}\right) | Helped |
Closest papers by meaning
- A general approximation lower bound in $L^p$ norm, with applications to feed-forward neural networks (NeurIPS 2022 · similarity 0.89)
- Generalization Bounds for Rank-sparse Neural Networks (NeurIPS 2025 · similarity 0.89)
- Fantastic Generalization Measures are Nowhere to be Found (ICLR 2024 · similarity 0.89)
- Generalizability of Neural Networks Minimizing Empirical Risk Based on Expressive Power (ICLR 2025 · similarity 0.89)
- Generalization Bounds and Model Complexity for Kolmogorov–Arnold Networks (ICLR 2025 · similarity 0.88)
- Koopman-based generalization bound: New aspect for full-rank weights (ICLR 2024 · similarity 0.88)
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.