AI Paper Atlas

Research map / Architectures and efficiency / Quantization

Improving Neural Network Efficiency via Post-Training Quantization With Adaptive Floating-Point

Fangxin Liu, Wenbo Zhao, Zhezhi He, Yanzhi Wang, Zongwu Wang, Changzhi Dai, Xiaoyao Liang, Li Jiang

ICCV 2021 · Architectures and efficiency › Quantization · approach: Low-precision training formats · 69 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 dynamic instead of static activation configuration (4.8-bit) (ablation)AFP ResNet-50 · ImageNet valtop-1 accuracy (%): 76.00 -> 76.09Improved
use dynamic instead of static activation configuration (3.9-bit) (ablation)AFP ResNet-50 · ImageNet valtop-1 accuracy (%): 75.11 -> 75.27Improved
use dynamic instead of static activation configuration (5.7-bit) (ablation)AFP MobileNet-v2 · ImageNet valtop-1 accuracy (%): 71.47 -> 71.59Improved
use dynamic instead of static activation configuration (4.8-bit) (ablation)AFP MobileNet-v2 · ImageNet valtop-1 accuracy (%): 70.91 -> 71.11Improved
reduce AFP bit-width from 4.8 to 3.9 bits (dynamic) (ablation)AFP ResNet-50 · ImageNet valtop-1 accuracy (%): 76.09 -> 75.27Hurt
reduce AFP bit-width from 5.7 to 4.8 bits (dynamic) (ablation)AFP MobileNet-v2 · ImageNet valtop-1 accuracy (%): 71.59 -> 71.11Hurt
search AFP configuration with Bayesian optimization instead of enumeration (ablation)AFP bit-width search · MobileNet-v2 and ResNet-50search iterations: 35 -> 18 iterations, 48.8% less timeImproved
use AFP over fixed 5-bit-style integer baselines at same bit-width (vs INT8)AFP dynamic vs INT8 · ImageNet val ResNet-50top-1 accuracy (%): 74.9 -> 76.09Improved
replace conventional quantizer/MAC with AFP shift-and-add hardwareAFP-encoded models · 90 nm synthesisenergy: 9.3x for MAC, 13.2x for quantizerImproved
increase calibration batch size up to 32 (ablation)AFP static quantization · ImageNet valtop-1 accuracy: accuracy stabilizes at batch size 32Improved
raise lambda from 0.25 to 0.50 in search objective (ablation)AFP search · MobileNet-v2 ImageNetchosen bit-width Nq: Nq 7 -> 5Improved

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