AI Paper Atlas

Research map / Optimization / Implicit bias and generalization

Impact of classification difficulty on the weight matrices spectra in Deep Learning and application to early-stopping

Xuran Meng, Jeff Yao

JMLR 2023 · Optimization › Implicit bias and generalization · approach: Stability-based generalization analysis · 5 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 spectral-criterion early stopping (FC2)NN1 · Gaussian D1 TP=0.15test accuracy: 20.17% -> 25.84%Improved
add spectral-criterion early stopping (FC2)NN1 · Gaussian D1 TP=0.6test accuracy: 88.30% -> 88.52%No clear effect
add spectral-criterion early stopping (FC1)NN2 · Gaussian D1 TP=0.13test accuracy: 87.50% -> 87.03%No clear effect
add spectral-criterion early stopping (FC2)LeNet · CIFAR10 batch 256test accuracy: 60.49% -> 54.23%Hurt
add spectral-criterion early stopping (FC1)MiniAlexNet · CIFAR10 batch 16test accuracy: 10%(explode) -> 69.05%Improved
add spectral-criterion early stopping (FC2)LeNet · MNIST batch 16test accuracy: 99.17% -> 99.08%No clear effect
raise threshold constant from C=0.4 to C=0.6 (ablation)spectral criterion · NN2 D1 syntheticstopping epoch found: epoch 6 at C=0.4 -> - at C=0.6 (TP=0.02, FC1)Mixed
decrease signal-to-noise ratio of Gaussian data (ablation)NN1/NN2 · Gaussian K-class dataterminal spectrum type: Strong SNR: Light Tail (MP Law); Weak SNR: Heavy TailHurt
increase number of classes K (ablation)NN1/NN2 · Gaussian K-class dataheavy tail emergence: K=2 -> K=8: HT more likely at same SNRHurt
train on CIFAR10 instead of MNIST (ablation)LeNet · batch 16spectrum type (detection rate): LT (99%) -> HT (64%)Hurt
increase batch size (ablation)VGG11 · CIFAR10spectrum type: spectra transit from BT to LT with increment of batch sizeImproved
go to narrower deeper layers (FC2 to FC4) (ablation)NN1 · Gaussian datatail weight: FC2 -> FC4: tails of spectrum bulks become lighterImproved

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