Research map / Optimization / Implicit bias and generalization
Non-Vacuous Generalisation Bounds for Shallow Neural Networks
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 a coupled prior trained on a prefix of the training data instead of a full-data-independent prior | SHEL with momentum SGD · Bin-M | generalisation bound: 0.837 -> 0.286 | Helped |
| Use a coupled prior trained on a prefix of the training data instead of a full-data-independent prior | SHEL with momentum SGD · Bin-F | generalisation bound: 0.426 -> 0.297 | Helped |
| Use a coupled prior trained on a prefix of the training data instead of a full-data-independent prior | SHEL with momentum SGD · MNIST | generalisation bound: 0.772 -> 0.490 | Helped |
| Use a coupled prior trained on a prefix of the training data instead of a full-data-independent prior | SHEL with momentum SGD · Fashion-MNIST | generalisation bound: 0.984 -> 0.727 | Helped |
| Use a coupled prior trained on a prefix of the training data instead of a full-data-independent prior | GELU with momentum SGD · MNIST | generalisation bound: 0.693 -> 0.293 | Helped |
| Use a coupled prior trained on a prefix of the training data instead of a full-data-independent prior | GELU with momentum SGD · Fashion-MNIST | generalisation bound: 0.976 -> 0.568 | Helped |
| Use a coupled prior trained on a prefix of the training data instead of a full-data-independent prior | SHEL without momentum SGD · Bin-M | generalisation bound: 0.835 -> 0.286 | Helped |
| Use a coupled prior trained on a prefix of the training data instead of a full-data-independent prior | SHEL without momentum SGD · Bin-F | generalisation bound: 0.425 -> 0.300 | Helped |
| Use a coupled prior trained on a prefix of the training data instead of a full-data-independent prior | SHEL without momentum SGD · MNIST | generalisation bound: 0.821 -> 0.522 | Helped |
| Use a coupled prior trained on a prefix of the training data instead of a full-data-independent prior | SHEL without momentum SGD · Fashion-MNIST | generalisation bound: 1.109 -> 0.844 | Helped |
| Use a coupled prior trained on a prefix of the training data instead of a full-data-independent prior | GELU without momentum SGD · MNIST | generalisation bound: 0.742 -> 0.317 | Helped |
| Use a coupled prior trained on a prefix of the training data instead of a full-data-independent prior | GELU without momentum SGD · Fashion-MNIST | generalisation bound: 1.100 -> 0.709 | Helped |
Closest papers by meaning
- Generalization Bounds for Neural Networks with Sparse Connectivity (NeurIPS 2026 · similarity 0.90)
- Generalization Bounds for Rank-sparse Neural Networks (NeurIPS 2025 · similarity 0.88)
- Generalizability of Neural Networks Minimizing Empirical Risk Based on Expressive Power (ICLR 2025 · similarity 0.88)
- Fantastic Generalization Measures are Nowhere to be Found (ICLR 2024 · similarity 0.87)
- Generalization Bounds and Model Complexity for Kolmogorov–Arnold Networks (ICLR 2025 · similarity 0.87)
- Stable Minima of ReLU Neural Networks Suffer from the Curse of Dimensionality: The Neural Shattering Phenomenon (NeurIPS 2025 · similarity 0.86)
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.