AI Paper Atlas

Research map / Architectures and efficiency / Quantization

Differentiable Joint Pruning and Quantization for Hardware Efficiency

Ying Wang, Yadong Lu, Tijmen Blankevoort

ECCV 2020 · Architectures and efficiency › Quantization · approach: Mixed-precision quantization · 83 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
optimize pruning and quantization jointly instead of in two stages (prune then quantize) (ablation)DJPQ on ResNet18 · ImageNettop-1 accuracy (%): 68.52 -> 69.27Improved
optimize pruning and quantization jointly instead of in two stages (prune then quantize) (ablation)DJPQ on ResNet18 · ImageNetBOP compression ratio: 46.53 -> 52.94Improved
add joint pruning to mixed-precision quantization (DJPQ vs DQ)mixed-precision quantization on ResNet18 · ImageNetBOPs (G): 40.71G vs 30.87G BOPsImproved
add joint pruning to mixed-precision quantization (DJPQ vs DQ)mixed-precision quantization on MobileNetV2 · ImageNetBOPs (G): 175.24G vs 132.28G BOPsImproved
learn power-of-two bit-widths during training instead of rounding DQ bit-widths afterwardsmixed-precision quantization on ResNet18 · ImageNetBOP compression ratio: 31.59 -> 52.28Improved
naively round DQ learned bit-widths up to powers of twoDQ on ResNet18 · ImageNetBOP compression ratio: 45.53 -> 31.59Hurt
restrict bit-widths to powers of two within DJPQDJPQ on VGG7 · CIFAR10top-1 accuracy (%): 91.54 -> 91.43No clear effect
restrict bit-widths to powers of two within DJPQDJPQ on ResNet18 · ImageNettop-1 accuracy (%): 69.27 -> 69.12No clear effect
use ADAM instead of SGD with manually scaled learning ratesDJPQ training · ImageNetaccuracyHurt

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