AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Sparse Winning Tickets are Data-Efficient Image Recognizers

Mukund Varma T, Xuxi Chen, Zhenyu Zhang, Tianlong Chen, Subhashini Venugopalan, Zhangyang Wang

NeurIPS 2022 · Architectures and efficiency › Pruning and efficient tuning · approach: Lottery ticket discovery · 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
replace dense network with IMP winning ticket, both with best augmentationResNet-18 · CIFAR10 (2%)accuracy: 55.14% -> 70.05%Improved
replace dense network with IMP winning ticket, both with best augmentationResNet-18 · CIFAR10 (1%)accuracy: 43.8% -> 59.66%Improved
add IMP winning ticket to ImageNet-initialized fine-tuningResNet-18 ImageNet init · CIFAR10 (1%)accuracy: 66.00% -> 69.46%Improved
add IMP winning ticket to SimCLR-initialized fine-tuningResNet-18 SimCLR init · CIFAR10 (2%)accuracy: 52.58% -> 64.09%Improved
add IMP to t-vMF losst-vMF loss · CLaMMaccuracy: 24.50% -> 59.67%Improved
add IMP to harmonic netsHarmonic nets · CIFAR10 (1%)accuracy: 22.97% -> 49.85%Improved
add IMP to ImageNet-initialized ResNet-18ResNet-18 ImageNet init · CLaMMaccuracy: 47.46% -> 55.86%Improved
add IMP winning ticket on small-sample datasets with Basic augmentationResNet-18 · EuroSATaccuracy: 83.44% -> 87.85%Improved
winning ticket on more complex many-class subsetsResNet-18 · CIFAR100 (2%)accuracy: 17.21% -> 25.06%Improved
winning ticket on ImageNet subsetResNet-18 · ImageNet (5%)accuracy: 28.82% -> 31.04%Improved
winning ticket on long-tailed dataResNet-18 · Long-tailed CIFAR100 (imbalance 0.01)accuracy: 45.13 -> 46.73Improved
winning ticket on mildly imbalanced dataResNet-18 · Long-tailed CIFAR100 (imbalance 0.10)accuracy: 64.77 -> 64.92No 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