AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Can Neural Network Memorization Be Localized?

Pratyush Maini, Michael Curtis Mozer, Hanie Sedghi, Zachary Chase Lipton, J Zico Kolter, Chiyuan Zhang

ICML 2023 · Architectures and efficiency › RNNs and normalization · approach: Representation geometry and collapse · 1 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
add example-tied dropout (memorization neurons dropped at test time)ResNet-9 · CIFAR-10 10% label noisenoisy train accuracy after dropout: 99.3% -> 3.1%Improved
add example-tied dropout (memorization neurons dropped at test time)ResNet-9 · MNIST 10% label noisenoisy train accuracy after dropout: 100% -> 0.1%Improved
add example-tied dropout (memorization neurons dropped at test time)ResNet-9 · SVHN 10% label noisenoisy train accuracy after dropout: 99.6% -> 1.4%Improved
add example-tied dropout: clean-example accuracy cost after memorization neuron dropResNet-9 · CIFAR-10 10% label noiseclean train accuracy: 99.9% -> 90.8%Hurt
add example-tied dropout: test accuracy after memorization neuron dropResNet-9 · CIFAR-10 10% label noisetest accuracy: 79.3% -> 82.7%Mixed
grid over p_mem and p_gen for example-tied dropout (ablation)ResNet-9 · CIFAR-10 10% label noisenoisy train accuracy after dropout: 2.8% to 6.9% across gridMixed
use standard dropout (p=0.4) instead of example-tied dropout (ablation)ResNet-9 · CIFAR-10 10% label noisenoisy train accuracy: 99.5% vs 3.10% for example-tiedHurt
use neuron-removal count as mislabeled-example detector instead of learning time or forgetting timeResNet-9 · CIFAR-10 10% label noise detectionAUC: 0.89 vs 0.87 (learning time) and 0.93 (forgetting time)Mixed
retrain rewound layers on clean examples only (layer retraining) vs. memorization in last layersResNet-9 · CIFAR-10 and SVHN 10% label noisenoisy example accuracy after clean-only retrainingNo clear effect
rewind final layers to earlier checkpoints (ablation)ResNet-9 / ResNet-50 / ViT · CIFAR-10 noisy training examplesaccuracy on memorized examplesNo clear effect

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