AI Paper Atlas

Research map / Architectures and efficiency / Quantization

HAWQ-V2: Hessian Aware trace-Weighted Quantization of Neural Networks

Zhen Dong, Zhewei Yao, Daiyaan Arfeen, Amir Gholami, Michael W. Mahoney, Kurt Keutzer

NeurIPS 2020 · Architectures and efficiency › Quantization · approach: Mixed-precision quantization · 54 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
use average Hessian trace rather than parameter-perturbation L2 alone to rank quantization sensitivity (ablation)HAWQ-V2 · ImageNet, SqueezeNextTop-1 accuracy: 67.83 -> 68.68Improved
choose the candidate precision setting with minimum total perturbation (ablation)HAWQ-V2 · ImageNet, SqueezeNextTop-1 accuracy: 67.46 -> 68.68Improved
replace HAWQ's top-eigenvalue sensitivity with trace-weighted automatic precision selectionHAWQ · ImageNet, ResNet50Top-1 accuracy: 75.48 -> 75.92Improved
replace manual mixed-precision selection with HAWQ-V2 automatic selectionHAWQ · ImageNet, InceptionV3Top-1 accuracy: 75.52 -> 75.98Improved
use trace-weighted automatic mixed precision instead of HAWQHAWQ · ImageNet, SqueezeNextTop-1 accuracy: 68.02 -> 68.68Improved
use HAWQ-V2 rather than FQN for quantizationRetinaNet-ResNet50 · Microsoft COCO 2017mAP: 32.5 -> 34.1Improved
use HAWQ-V2 rather than direct uniform quantizationRetinaNet-ResNet50 · Microsoft COCO 2017mAP: 31.5 -> 34.1Improved
use trace-weighted HAWQ-V2 rather than HAWQRetinaNet-ResNet50 · Microsoft COCO 2017mAP: 33.5 -> 34.1Improved
use mixed-precision rather than uniform 4-bit activationsHAWQ-V2 · Microsoft COCO 2017mAP: 34.1 -> 34.4Improved
increase activation precision from 4-bit to 6-bitHAWQ-V2 · Microsoft COCO 2017mAP: 34.1 -> 34.8Improved
use mixed-precision quantization rather than direct quantizationRetinaNet-ResNet50 · Microsoft COCO 2017mAP: 31.5 -> 34.1Improved

Later papers that compared with this method

4 later accepted papers report HAWQv2 as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.

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