AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Utilizing Class Information for Deep Network Representation Shaping

Daeyoung Choi, Wonjong Rhee

AAAI 2019 · Label-efficient and robust learning › Recognition and long-tail · approach: Loss design and classifier theory · 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
Apply class-wise variance regularizationCNN · CIFAR-10, Momentumerror (%): 25.78 ± 0.37 -> 24.42 ± 0.31Improved
Apply class-wise variance regularizationResNet-32 · CIFAR-10error (%): 7.39 -> 7.17Improved
Apply class-wise variance regularizationResNet-110 · CIFAR-10error (%): 6.12 -> 6.00Improved
Apply class-wise variance regularization6-layer autoencoder · MNIST autoencodermean squared error: 1.44 × 10^{-2} ± 3.36 × 10^{-4} -> 1.19 × 10^{-2} ± 2.48 × 10^{-4}Improved
Use class-wise covariance regularization instead of class-wise variance regularization (ablation)6-layer MLP · MNIST MLP, defaulterror (%): 2.42 ± 0.06 -> 2.49 ± 0.10Hurt
Remove class conditioning from variance regularization (ablation)6-layer MLP · MNIST MLP, defaulterror (%): 2.42 ± 0.06 -> 2.65 ± 0.11Hurt
Apply all-class variance regularization rather than no regularizer (ablation)CNN · CIFAR-10, Adamerror (%): 26.64 ± 0.16 -> 21.44 ± 0.88Improved
Move representation regularization from lower layers toward upper layers (ablation)deep networks · MNIST MLP and CIFAR-10 CNNerror (%)Mixed

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