Research map / Optimization / Implicit bias and generalization
Free Probability for predicting the performance of feed-forward fully connected neural networks
Reda CHHAIBI, Tariq Daouda, Ezechiel Kahn
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 the 90th Jacobian-spectrum quantile as a predictor of final test accuracy | Randomly sampled 4-layer ReLU MLPs · MNIST | R factor: 0.84 | Improved |
| Use the 90th Jacobian-spectrum quantile as a predictor of final test accuracy | Randomly sampled 4-layer ReLU MLPs · FashionMNIST | R factor: 0.81 | Improved |
| Use the 90th Jacobian-spectrum quantile as a predictor of final test accuracy | Randomly sampled 4-layer ReLU MLPs · CIFAR10 | R factor: 0.75 | Improved |
| Use all Jacobian-spectrum quantiles instead of only the 90th quantile | Accuracy regression for randomly sampled 4-layer ReLU MLPs · MNIST | R factor: 0.84 -> 0.89 | Improved |
| Use all Jacobian-spectrum quantiles instead of only the 90th quantile | Accuracy regression for randomly sampled 4-layer ReLU MLPs · FashionMNIST | R factor: 0.81 -> 0.87 | Improved |
| Use all Jacobian-spectrum quantiles instead of only the 90th quantile | Accuracy regression for randomly sampled 4-layer ReLU MLPs · CIFAR10 | R factor: 0.75 -> 0.83 | Improved |
| Use higher rather than lower Jacobian-spectrum quantiles as a single accuracy predictor | Randomly sampled 4-layer ReLU MLPs · MNIST | R factor: 0.058 -> 0.845 | Improved |
| Use higher rather than lower Jacobian-spectrum quantiles as a single accuracy predictor | Randomly sampled 4-layer ReLU MLPs · FashionMNIST | R factor: 0.035 -> 0.807 | Improved |
| Use higher rather than lower Jacobian-spectrum quantiles as a single accuracy predictor | Randomly sampled 4-layer ReLU MLPs · CIFAR10 | R factor: -0.016 -> 0.755 | Improved |
Closest papers by meaning
- Spectral Asymptotics of Neural Network Jacobians: Convergency, Universality, and Phase Transition (NeurIPS 2026 · similarity 0.88)
- Free Decompression with Algebraic Spectral Curves (NeurIPS 2026 · similarity 0.87)
- An Investigation into Neural Net Optimization via Hessian Eigenvalue Density (ICML 2019 · similarity 0.86)
- An Infinite-Width Analysis on the Jacobian-Regularised Training of a Neural Network (ICML 2024 · similarity 0.86)
- Convergence Analysis of Newton's Method for Neural Networks in the Overparameterized Limit (NeurIPS 2026 · similarity 0.85)
- High-dimensional Limit of SGD for Diagonal Linear Networks (NeurIPS 2026 · similarity 0.85)
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.
