AI Paper Atlas

Research map / Architectures and efficiency / Quantization

Learning Compression from Limited Unlabeled Data

Xiangyu He, Jian Cheng

ECCV 2018 · Architectures and efficiency › Quantization · approach: Post-training weight quantization · 29 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 BN re-normalization after 4-bit quantization (ablation)ResNet-18 · ImageNet val 4-bit top-1top-1 accuracy gap: -13.21 -> -1.83Improved
add BN re-normalization after 4-bit quantization (ablation)MobileNet · ImageNet val 4-bit top-1top-1 accuracy gap: -70.80 -> -9.75Improved
add BN re-normalization after 4-bit quantization (ablation)ResNet-50 · ImageNet val 4-bit top-1top-1 accuracy gap: -7.16 -> -2.14Improved
add BN re-normalization after 4-bit quantization (ablation)AlexNet · ImageNet val 4-bit top-1top-1 accuracy gap: -1.77 -> -0.39Improved
add BN re-normalization after 4-bit quantization (ablation)AlexNet · ImageNet val 4-bit top-5top-5 accuracy gap: -1.30 -> -0.20Improved
add BN re-normalization after weight pruning (ablation)MobileNet · ImageNet val 3x pruningtop-5 accuracy: +53.82% Top-5 improvement to 78.43%Improved
add BN re-normalization after weight pruning (ablation)ResNet-50 · ImageNet val 3x pruningtop-5 accuracy: +6.92% Top-5 improvement to 90.00%Improved
add BN re-normalization to 3-bit weight AlexNet (ablation)AlexNet · ImageNet val 3-8-bittop-1 accuracy: +25.43% Top-1 to 50.69%Improved
use 2^n quantizer instead of round-to-nearestAlexNet 4-bit weights · ImageNet val top-1top-1 accuracy: 48.32 -> 58.66Improved
use 2^n quantizer instead of uniformResNet-18 4-bit weights · ImageNet val top-1top-1 accuracy: 50.18 -> 55.76Improved
use KL divergence metrics instead of l2 for weight quantizationAlexNet 4-bit 2^n · ImageNet val top-1top-1 accuracy: 58.90 -> 56.30Hurt
add feature-map fitting term to weight quantization objectiveQuasi-Lloyd-Max · ImageNet valaccuracyNo 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