AI Paper Atlas

Research map / Architectures and efficiency / Quantization

Weight Fixing Networks

Christopher Subia-Waud, Srinandan Dasmahapatra

ECCV 2022 · Architectures and efficiency › Quantization · approach: Binary neural networks · 1 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
WFN instead of APoT 3-bit (unique weights)ResNet-18 · ImageNetunique parameter count: 9237 -> 164Improved
WFN delta=0.01 instead of APoT (weight entropy)ResNet-18 · ImageNetweight-space entropy: 5.77 -> 3.01Improved
WFN delta=0.01 instead of APoT (weight entropy)ResNet-34 · ImageNetweight-space entropy: 6.77 -> 3.48Improved
WFN delta=0.0075 vs APoT top-1ResNet-18 · ImageNettop-1 accuracy: 69.9 -> 70.3Improved
WFN delta=0.0075 vs baselineGoogLeNet · ImageNettop-1 accuracy: 69.7 -> 70.9Improved
larger delta (0.0075 -> 0.015) in WFN (ablation)ResNet-18 · ImageNettop-1 accuracy: 70.3 -> 67.3Hurt
larger delta (0.0075 -> 0.015) in WFN (ablation)ResNet-34 · ImageNettop-1 accuracy: 73.0 -> 72.2Hurt
larger delta (0.0075 -> 0.015) in WFN (ablation)ResNet-18 · ImageNetunique parameter count: 193 -> 90Improved
WFN delta=0.015 vs APoT compressionResNet-18 · ImageNetcompression ratio: 10.2 -> 13.4Improved
WFN delta=0.0075 vs LQ-NetResNet-50 · ImageNettop-1 accuracy: 74.2 -> 76.0Improved
WFN delta=0.015 vs baselineResNet-34 · ImageNettop-1 accuracy: 73.3 -> 72.2Hurt
relative-distance noise instead of absolute noise to weights (ablation)pre-trained ResNet-18 · ImageNet noise testaccuracy toleranceImproved

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