AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Compressing Models With Few Samples: Mimicking Then Replacing

Huanyu Wang, Junjie Liu, Xin Ma, Yang Yong, Zhenhua Chai, Jianxin Wu

CVPR 2022 · Architectures and efficiency › Pruning and efficient tuning · approach: Joint pruning and quantization · 11 citations (OpenAlex)

Check 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
Mimic features before rather than after final pooling (ablation)MiR on ResNet-34 · ILSVRC-2012, Prune-C Normal, 500 samplestop-1 accuracy: 65.4 -> 67.0Improved
Mimic features before rather than after final pooling (ablation)MiR on ResNet-34 · ILSVRC-2012, Prune-C Residual, 500 samplestop-1 accuracy: 62.4 -> 65.4Improved
Use holistic feature matching instead of layer-wise reconstructionPruned ResNet-34 · ILSVRC-2012, Prune-C Normal, 500 samplestop-1 accuracy: CD 63.7 -> MiR_before 67.0Improved
Use holistic feature matching instead of layer-wise reconstructionPruned ResNet-34 · ILSVRC-2012, Prune-C Residual, 500 samplestop-1 accuracy: FSKD 42.8 -> MiR_before 65.4Improved
Use MiR feature mimicking on MobileNetV2 rather than label-based fine-tuningPruned MobileNetV2 · ILSVRC-2012, Prune-D, 500 samplestop-1 accuracy: KD 48.4 -> MiR_before 67.6Improved
Use MiR_before rather than CDPruned ResNet-56 · CIFAR-10, 1 sample per classtop-1 accuracy: 89.00 -> 89.27Improved
Replace MSE with L1 feature loss (ablation)MiR on ResNet-34 · ILSVRC-2012, Prune-B CD-style, 500 samplestop-1 accuracy: 69.75 -> 69.88No clear effect
Replace MSE with cosine similarity loss (ablation)MiR on ResNet-34 · ILSVRC-2012, Prune-B CD-style, 500 samplestop-1 accuracy: 69.75 -> 69.07Hurt
Replace MSE with LSH loss (ablation)MiR on ResNet-34 · ILSVRC-2012, Prune-B CD-style, 500 samplestop-1 accuracy: 69.75 -> 66.89Hurt
Add L1 loss to MSE (ablation)MiR on ResNet-34 · ILSVRC-2012, Prune-B CD-style, 500 samplestop-1 accuracy: 69.75 -> 69.89No clear effect
Tune the classifier after mimicking, while freezing the trained backbone (ablation)MiR · ILSVRC-2012, Prune-B CD-style, 500 samplestop-1 accuracy: 69.75 -> 66.44 (LR 0.1); 69.75 -> 69.96 (LR 0.05)Mixed
Tune the classifier while freezing an untrained backbone (ablation)Pruned ResNet-34 · ILSVRC-2012, Prune-B CD-style, 500 samplestop-1 accuracy: 69.75 -> 45.43 (LR 0.1)Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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