AI Paper Atlas

Research map / Architectures and efficiency / Quantization

RDO-Q: Extremely Fine-Grained Channel-Wise Quantization via Rate-Distortion Optimization

Zhe Wang, Jie Lin, Xue Geng, Mohamed M. Sabry Aly, Vijay Chandrasekhar

ECCV 2022 · Architectures and efficiency › Quantization · approach: Mixed-precision quantization · 8 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 RDO-Q channel-wise allocation vs best prior 2-bit methodResNet-18 · ImageNet 2-bittop-1 accuracy (%): 67.8% -> 68.8%Improved
use RDO-Q vs best prior 2-bit methodResNet-34 · ImageNet 2-bittop-1 accuracy (%): 72.1% -> 72.6%Improved
use RDO-Q vs best prior 2-bit methodResNet-50 · ImageNet 2-bittop-1 accuracy (%): 73.8% -> 75.0%Improved
use RDO-Q vs SAT at 4 bitsMobileNet-v2 · ImageNet 4-bittop-1 accuracy (%): 71.1% -> 71.3%Improved
add hardware-aware constraints, ResNet-50 vs uncompressed (ablation)ResNet-50 · Google TPUinference rate (images/s): 361 -> 1254Improved
add hardware-aware constraints, ResNet-50 vs uncompressed (ablation)ResNet-50 · MIT Eyerissinference rate (images/s): 5.6 -> 17.0Improved
hardware-aware RDO-Q vs equal-bit DoReFa+PACT at 76.5% accuracyResNet-50 · Google TPUinference rate (images/s): 646 -> 769Improved
hardware-aware RDO-Q vs HAQMobileNet-V2 · Google TPUinference rate (images/s): 2067 -> 2197Improved
rate-distortion search instead of reinforcement learning (HAQ)ResNet-50 · bit allocation search timetime: 117 hours -> 7 minitesImproved
rate-distortion search instead of reinforcement learning (HAQ)MobileNet-v2 · bit allocation search timetime: 79 hours -> 6 minitesImproved
use channel-wise instead of layer-wise bit allocation (ablation)MobileNet-V2 · ImageNettop-1 accuracy (%)Improved
add hardware-aware constraints to channel-wise allocation (ablation)MobileNet-V2 · Eyeriss and TPUinference rateImproved

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