AI Paper Atlas

Research map / Architectures and efficiency / Quantization

Mixed-Precision Neural Network Quantization via Learned Layer-Wise Importance

Chen Tang, Kai Ouyang, Zhi Wang, Yifei Zhu, Wen Ji, Yaowei Wang, Wenwu Zhu

ECCV 2022 · Architectures and efficiency › Quantization · approach: Mixed-precision quantization · 97 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
assign more bits to layers with larger learned scale factors (not reversed) (ablation)MobileNetV1 W4MP/A4MP · ImageNet 9.68G BitOpstop-1 accuracy (%): 65.25 -> 71.84Improved
use importance-guided assignment instead of reversed at 3-bit level versus reversed 4-bit (ablation)MobileNetV1 · ImageNettop-1 accuracy (%): 65.25 -> 69.48Improved
initialize from a higher-accuracy FP model (70.5 vs 69.6)Ours ResNet-18 W3MP/A3MP · ImageNettop-1 accuracy (%): 69.0 -> 69.7Improved
initialize from a higher-accuracy FP model (70.5 vs 69.6)Ours ResNet-18 W4MP/A4MP · ImageNettop-1 accuracy (%): 70.1 -> 70.8Improved
tighten constraint from 3W3A to 2.5W3A (19.81G BitOps)Ours ResNet-18 · ImageNettop-1 accuracy (%): 69.0 -> 68.7Hurt
use learned-importance MPQ instead of DNAS at 3-bitResNet-18 · ImageNettop-1 accuracy (%): 68.7 -> 69.7Improved
use learned-importance MPQ instead of HAWQv2 (ILP search)ResNet-50 · ImageNettop-1 accuracy (%): 75.8 -> 76.9Improved
use learned-importance MPQ instead of HAQ (3MP weights)MobileNetV1 weight-only · ImageNettop-1 accuracy (%): 67.66 -> 71.57Improved
use learned-importance MPQ instead of HAQ (4MP weights)MobileNetV1 weight-only · ImageNettop-1 accuracy (%): 71.74 -> 72.60Improved
use learned-importance MPQ instead of SPOS at 4-bitResNet-18 · ImageNettop-1 accuracy (%): 70.5 -> 70.8Improved
use ILP search on learned importance instead of AutoQ searchResNet-50 search time · policy search costsearch time: 1000z GPU-hours -> 330z x speedupImproved

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