AI Paper Atlas

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

Deep Subspace Clustering with Data Augmentation

Mahdi Abavisani, Alireza Naghizadeh, Dimitris Metaxas, Vishal Patel

NeurIPS 2020 · Label-efficient and robust learning › Recognition and long-tail · approach: Data augmentation and generalization bias

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 data augmentation only (no consistency loss) (ablation)DSC · Extended Yale Bclustering error (%): 2.67 -> 3.10Hurt
add data augmentation only (no consistency loss) (ablation)MLRDSC · Extended Yale Bclustering error (%): 1.36 -> 2.84Hurt
add consistency loss without augmentation (ablation)DSC · Extended Yale Bclustering error (%): 2.67 -> 2.56Improved
add consistency loss without augmentation (ablation)MLRDSC · Extended Yale Bclustering error (%): 1.36 -> 0.95Improved
add augmentation together with consistency loss (ablation)DSC · Extended Yale Bclustering error (%): 2.56 -> 1.92Improved
add augmentation together with consistency loss (ablation)MLRDSC · Extended Yale Bclustering error (%): 0.95 -> 0.82Improved
use MLRDSC backbone instead of DSC (ablation)MLRDSC-DA · Extended Yale Bclustering error (%): 1.92 -> 0.82Improved
use searched policies instead of Random LR Flips (ablation)MLRDSC-DA · Extended Yale Bclustering error (%): 1.32 -> 0.82Improved
use searched policies instead of Cut-out (ablation)MLRDSC-DA · Extended Yale Bclustering error (%): 2.88 -> 0.82Improved
use searched policies instead of common practitioner augmentations (ablation)MLRDSC-DA · Extended Yale Bclustering error (%): 2.96 -> 0.82Improved
use searched policies instead of AutoAugment ImageNet policy (ablation)MLRDSC-DA · Extended Yale Bclustering error (%): 5.96 -> 0.82Improved
use searched policies instead of AutoAugment SVHN policy (ablation)MLRDSC-DA · Extended Yale Bclustering error (%): 11.31 -> 0.82Improved

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