AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Get More at Once: Alternating Sparse Training with Gradient Correction

Li Yang, Jian Meng, Jae-sun Seo, Deliang Fan

NeurIPS 2022 · Architectures and efficiency › Pruning and efficient tuning · approach: Sparse training from scratch · 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 gradient correction (GC) to AST (si=0%) (ablation)AST · CIFAR-10 wide ResNet-32, 98% sparsityTop-1 accuracy: 92.44 -> 92.69Improved
Add gradient correction (GC) to AST (si=90%) (ablation)AST · CIFAR-10 wide ResNet-32, 90% sparsityTop-1 accuracy: 92.32 -> 92.90Improved
Add gradient correction to AST (si=0%) (ablation)AST · CIFAR-100 wide ResNet-32, 90% sparsityTop-1 accuracy: 73.12 -> 73.41Improved
Add gradient correction to AST (si=50%) (ablation)AST · ImageNet ResNet-50, 80% prune ratioTop-1 accuracy: 72.6 -> 73.2Improved
Add gradient correction to AST (si=50%) (ablation)AST · ImageNet ResNet-50, 90% prune ratioTop-1 accuracy: 72.3 -> 73.1Improved
Use non-disjoint (ND) sub-nets instead of completely subset (CS) (ablation)AST · CIFAR-10 ResNet-32, 70% pruningAccuracy: 93.69 -> 94.47Improved
Use non-disjoint instead of completely subset sub-nets (ablation)AST · CIFAR-10 ResNet-32, 98% pruningAccuracy: 92.49 -> 92.76Improved
AST+GC instead of Jointly-Trained S-NetAST+GC · CIFAR-10 wide ResNet-32, 90% sparsityTop-1 accuracy: 92.59 -> 93.88Improved
AST+GC instead of Jointly-Trained S-NetAST+GC · CIFAR-100 wide ResNet-32, 90% sparsityTop-1 accuracy: 70.40 -> 73.41Improved
AST+GC instead of individually trained GraNet (3x cost)AST+GC · CIFAR-100 wide ResNet-32, 98% sparsityTop-1 accuracy: 69.89 -> 68.42Hurt
AST+GC instead of GraNetAST+GC · CIFAR-10 wide ResNet-32, 90% sparsityTop-1 accuracy: 94.12 -> 93.88Hurt
Train more sub-nets (2 to 5) with AST instead of individual training (ablation)AST · CIFAR-10 ResNet-18, 98% sparsityAccuracy: 94.50 -> 94.18Mixed

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