AI Paper Atlas

Research map / Architectures and efficiency / Quantization

Post-Training Piecewise Linear Quantization for Deep Neural Networks

Jun Fang, Ali Shafiee, Hamzah Abdel-Aziz, David Thorsley, Georgios Georgiadis, Joseph H. Hassoun

ECCV 2020 · Architectures and efficiency › Quantization · approach: Mixed-precision quantization

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
use piecewise linear instead of uniform quantization (4-bit weights)Inception-v3 · ImageNet valtop-1 accuracy (%): 44.28 -> 75.72Improved
use piecewise linear instead of uniform quantization (4-bit weights)ResNet-50 · ImageNet valtop-1 accuracy (%): 65.48 -> 74.28Improved
use piecewise linear instead of uniform quantization (4-bit weights)MobileNet-v2 · ImageNet valtop-1 accuracy (%): 11.37 -> 54.34Improved
use piecewise linear instead of uniform quantization (8-bit weights)Inception-v3 · ImageNet valtop-1 accuracy (%): 77.53 -> 77.52No clear effect
add bias correction to 4-bit PWLQ (ablation)MobileNet-v2 · ImageNet valtop-1 accuracy (%): 54.34 -> 69.22Improved
add bias correction to 4-bit PWLQ (ablation)Inception-v3 · ImageNet valtop-1 accuracy (%): 75.72 -> 76.45Improved
use two breakpoints instead of one (3-bit weights) (ablation)MobileNet-v2 · ImageNet valtop-1 accuracy (%): 16.77 -> 36.44Improved
use three breakpoints instead of one (3-bit weights) (ablation)MobileNet-v2 · ImageNet valtop-1 accuracy (%): 16.77 -> 55.17Improved
use two breakpoints instead of one (4-bit weights) (ablation)ResNet-50 · ImageNet valtop-1 accuracy (%): 74.28 -> 75.24Improved
perturb optimal breakpoints by 5% (ablation)4-bit Inception-v3 · ImageNet valtop-1 accuracy (%): 75.72 -> 74.05Hurt
use PWLQ instead of uniform quantization (4-bit weights)DeepLab-v3+ · Pascal VOC segmentationmIoU (%): 50.05 -> 67.66Improved
use PWLQ instead of uniform quantization (4-bit weights)SSD-Lite · Pascal VOC detectionmAP (%): 64.79 -> 68.32Improved

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