AI Paper Atlas

Research map / Architectures and efficiency / Quantization

Training wide residual networks for deployment using a single bit for each weight

Mark D. McDonnell

ICLR 2018 · Architectures and efficiency › Quantization · approach: Low-precision training formats · 41 citations (OpenAlex)

Read the paper

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 1-bit weights instead of full-precision weightsWRN 20-10 · CIFAR-10, standard augmentationerror rate: 4.22 -> 4.72Hurt
Use 1-bit weights instead of full-precision weightsWRN 20-10 · CIFAR-100, standard augmentationerror rate: 18.76 -> 19.35Hurt
Add cutout augmentation (ablation)1-bit WRN 20-10 · CIFAR-10error rate: 4.72 -> 3.92Improved
Add cutout augmentation (ablation)1-bit WRN 20-10 · CIFAR-100error rate: 19.35 -> 18.51Improved
Do not learn batch-normalization scale and offset parameters (ablation)Wide ResNets · CIFAR-100error rate: around 3% absoluteImproved
Do not learn batch-normalization scale and offset parameters (ablation)Wide ResNets · ImageNet32error rateHurt
Use warm-restart learning-rate schedule (ablation)1-bit Wide ResNets · CIFAR-100error rateImproved
Use warm-restart learning-rate schedule (ablation)Full-precision Wide ResNets · CIFAR-100error rateNo clear effect
Use warm-restart learning-rate schedule (ablation)Wide ResNets · CIFAR-10/100convergence: just 30 epochsImproved
Use 1-bit weights instead of full-precision weightsWRN 20-4 · SVHNerror rate: 1.75 -> 1.93Hurt
Use 1-bit weights instead of full-precision weightsWRN 20-15 · ImageNet32Top-1 error: 38.90 -> 41.26Hurt
Use 1-bit weights instead of full-precision weightsWRN 18-2.5 · ImageNet, single cropTop-1 error: 26.92 -> 31.03Hurt

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