AI Paper Atlas

Research map / Learning theory and algorithms / Kernels and regression theory

On Uniform Convergence and Low-Norm Interpolation Learning

Lijia Zhou, Danica J. Sutherland, Nati Srebro

NeurIPS 2020 · Learning theory and algorithms › Kernels and regression theory · approach: Information-theoretic generalization bounds · 10 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
Restrict the norm-bounded class to predictors that interpolate the training dataUniform convergence in junk-features Gaussian linear regression · Junk-features Gaussian linear regressionasymptotic worst-case excess risk: α^2 L_{\mathcal{D}}(w^*)Improved
Allow interpolators norm at most α times the minimum norm, with α ≥ 1Minimum-norm interpolation · Junk-features Gaussian linear regressionasymptotic worst-case excess risk: α^2 L_{\mathcal{D}}(w^*)Mixed
Use a norm ball containing the minimum-norm interpolator without requiring interpolationMinimum-norm interpolator · Junk-features Gaussian linear regressiontwo-sided uniform generalization gap: \inftyHurt
Use typical algorithm outputs in a two-sided uniform convergence boundSymmetric interpolation algorithms · Junk-features Gaussian linear regressiontwo-sided uniform generalization gap: 3 \sigma^2Hurt
Set the ridge parameter to λ_n = o(n)Minimum-norm interpolator · Junk-features Gaussian linear regressionexpected excess population risk: 0Improved
Use an interpolator with norm no more than a constant factor above the minimum normLow-norm interpolation · Junk-features Gaussian linear regressionasymptotic excess population risk: α^2 L_{\mathcal{D}}(w^*)Improved

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