AI Paper Atlas

Research map / Optimization / Implicit bias and generalization

Data Cleansing for Models Trained with SGD

Satoshi Hara, Atsushi Nitanda, Takanori Maehara

NeurIPS 2019 · Optimization › Implicit bias and generalization · approach: Stability-based generalization analysis · 32 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 SGD-step retracing to estimate influence instead of K&LInfluence estimation for LogReg and DNN · Adult, 20News, and MNISTKendall's tau: .93 (.02) vs .85 (.07); .94 (.05) vs .82 (.15); .95 (.02) vs .70 (.15) for LogReg; .75 (.10) vs .54 (.12); .45 (.12) vs Improved
Use SGD-step retracing to estimate influence instead of K&LInfluence estimation for LogReg and DNN · Adult, 20News, and MNISTJaccard index: .80 (.10) vs .60 (.17); .79 (.15) vs .52 (.19); .83 (.10) vs .41 (.16) for LogReg; .59 (.16) vs .32 (.11); .25 (.08) vs Improved
Remove the most influential instances identified by the proposed methodCNN trained on MNIST and CIFAR10 · MNIST and CIFAR10 test setsmisclassification rate: 25 cases in MNIST and 26 cases in CIFAR10 out of 30 experiments showed decreased misclassification ratesImproved
Retrace only one SGD epoch instead of all SGD steps (ablation)Proposed influence estimation for data cleansing · MNIST and CIFAR10misclassification rateNo clear effect
Remove instances ranked by Autoencoder outliernessCNN trained on MNIST and CIFAR10 · MNIST and CIFAR10 test setsmisclassification rateNo clear effect
Remove influential instances identified by the proposed method rather than by K&LCNN data cleansing · MNIST and CIFAR10misclassification rate: 10 and 1000; 10 and 100Improved

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