AI Paper Atlas

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

Strong inductive biases provably prevent harmless interpolation

Michael Aerni, Marco Milanta, Konstantin Donhauser, Fanny Yang

ICLR 2023 · Learning theory and algorithms › Kernels and regression theory · approach: Double descent and overfitting · 0 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 a small CNN filter size rather than a large oneCNNs on synthetic images with noisy labels · Synthetic images, 20% label noisetest error gap between interpolation and optimal early stoppingHurt
Increase CNN filter sizeCNNs on synthetic images with noisy labels · Synthetic images, 20% label noisetest error gap between interpolation and optimal early stoppingImproved
Interpolate instead of optimally early-stopCNNs on synthetic images with noiseless labels · Synthetic images, noiselesstest errorNo clear effect
Use a small filter size and optimal early stoppingCNNs on synthetic images with 20% label noise · Noisy training subsettraining 0-1 errorImproved
Use a large filter size and optimal early stoppingCNNs on synthetic images with 20% label noise · Noisy training subsettraining 0-1 errorHurt
Increase rotational-invariance augmentationWide ResNets on EuroSAT with noisy labels · EuroSAT, 20% label noisetest error gap between interpolation and optimal early stoppingHurt
Reduce the number of augmented rotationsWide ResNets on EuroSAT with noisy labels · EuroSAT, 20% label noisetest error gap between interpolation and optimal early stoppingImproved
Use rotational invariance and optimal early stoppingWide ResNets on EuroSAT with 20% label noise · Noisy training subsettraining 0-1 errorImproved
Use no rotational invariance and optimal early stoppingWide ResNets on EuroSAT with 20% label noise · Noisy training subsettraining 0-1 errorHurt
Interpolate rather than use optimal ridge regularizationConvolutional-kernel regression with a strong inductive bias (small filter size) · High-dimensional kernel regression with noisy datapopulation riskHurt
Interpolate rather than use optimal ridge regularizationConvolutional-kernel regression with a weak inductive bias (large filter size) · High-dimensional kernel regression with noisy datapopulation riskNo clear effect
Use a weak inductive bias and optimally tune ridge regularizationConvolutional-kernel regression · High-dimensional kernel regression with noisy datatraining error relative to noise levelImproved

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