AI Paper Atlas

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

NeurIPS 2022 · Optimization › Implicit bias and generalization · approach: Edge-of-stability dynamics · 2 citations (OpenAlex)

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 changedWhereMeasuredEffect
Use the 90th Jacobian-spectrum quantile as a predictor of final test accuracyRandomly sampled 4-layer ReLU MLPs · MNISTR factor: 0.84Improved
Use the 90th Jacobian-spectrum quantile as a predictor of final test accuracyRandomly sampled 4-layer ReLU MLPs · FashionMNISTR factor: 0.81Improved
Use the 90th Jacobian-spectrum quantile as a predictor of final test accuracyRandomly sampled 4-layer ReLU MLPs · CIFAR10R factor: 0.75Improved
Use all Jacobian-spectrum quantiles instead of only the 90th quantileAccuracy regression for randomly sampled 4-layer ReLU MLPs · MNISTR factor: 0.84 -> 0.89Improved
Use all Jacobian-spectrum quantiles instead of only the 90th quantileAccuracy regression for randomly sampled 4-layer ReLU MLPs · FashionMNISTR factor: 0.81 -> 0.87Improved
Use all Jacobian-spectrum quantiles instead of only the 90th quantileAccuracy regression for randomly sampled 4-layer ReLU MLPs · CIFAR10R factor: 0.75 -> 0.83Improved
Use higher rather than lower Jacobian-spectrum quantiles as a single accuracy predictorRandomly sampled 4-layer ReLU MLPs · MNISTR factor: 0.058 -> 0.845Improved
Use higher rather than lower Jacobian-spectrum quantiles as a single accuracy predictorRandomly sampled 4-layer ReLU MLPs · FashionMNISTR factor: 0.035 -> 0.807Improved
Use higher rather than lower Jacobian-spectrum quantiles as a single accuracy predictorRandomly sampled 4-layer ReLU MLPs · CIFAR10R factor: -0.016 -> 0.755Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy