AI Paper Atlas

Research map / Architectures and efficiency / Quantization

Extremely Low Bit Neural Network: Squeeze the Last Bit Out With ADMM

Cong Leng, Zesheng Dou, Hao Li, Shenghuo Zhu, Rong Jin

AAAI 2018 · Architectures and efficiency › Quantization · approach: Low-precision training formats · 289 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 ADMM ternary quantization instead of TWNAlexNet · ImageNet validationTop-1 accuracy: 0.575 -> 0.582Improved
Use ADMM ternary quantization instead of TWNVGG-16 · ImageNet validationTop-1 accuracy: 0.691 -> 0.700Improved
Use ADMM ternary quantization instead of TWNResNet-18 · ImageNet validationTop-1 accuracy: 0.618 -> 0.670Improved
Use ADMM ternary quantization instead of TWNResNet-50 · ImageNet validationTop-1 accuracy: 0.656 -> 0.725Improved
Use ADMM ternary quantization instead of TWNGoogleNet · ImageNet validationTop-1 accuracy: 0.612 -> 0.631Improved
Increase weight precision from binary to ternary (ablation)ImageNet CNNs · ImageNet validationTop-1 accuracyImproved
Quantize GoogleNet 1×1 kernels with INT8 instead of extremely low bits (ablation)GoogleNet · ImageNet validationTop-1 accuracy: 0.603 -> 0.654; 0.631 -> 0.667Improved
Quantize Darknet 1×1 kernels with INT8 instead of extremely low bits (ablation)Darknet+SSD · Pascal VOC 2007 testmAP: 0.609 (0.621); 0.624 (0.639)Improved
Replace full precision weights with {-4, +4} quantized weights (ablation)VGG-16+SSD · Pascal VOC 2007 testmAP: 0.778 -> 0.776Hurt

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