AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Training Binary Weight Networks via Semi-Binary Decomposition

Qinghao Hu, Gang Li, Peisong Wang, Yifan Zhang, Jian Cheng

ECCV 2018 · Architectures and efficiency › RNNs and normalization · approach: Efficient network compression and training · 18 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
minimize feature-map quantization loss (SBD-FQ) instead of direct weight decomposition (ablation)semi-binary decomposition · ImageNet-2012 val, AlexNettop-1 accuracy (%): 58.0 -> 58.5Improved
minimize feature-map quantization loss (SBD-FQ) instead of direct weight decomposition (ablation)semi-binary decomposition · ImageNet-2012 val, ResNet-18top-1 accuracy (%): 64.9 -> 66.2Improved
minimize feature-map quantization loss (SBD-FQ) instead of direct weight decomposition (ablation)semi-binary decomposition · ImageNet-2012 val, ResNet-50top-1 accuracy (%): 67.7 -> 68.9Improved
use semi-binary decomposition (SBD-FQ) instead of BWN scaling-factor binarizationbinary weight ResNet-18 · ImageNet-2012 val, ResNet-18top-1 accuracy (%): 60.8 -> 66.2Improved
use semi-binary decomposition (SBD-FQ) instead of BWNbinary weight ResNet-50 · ImageNet-2012 val, ResNet-50top-1 accuracy (%): 63.9 -> 68.9Improved
use semi-binary decomposition (SBD-FQ) instead of BWNbinary weight AlexNet · ImageNet-2012 val, AlexNettop-1 accuracy (%): 56.8 -> 58.5Improved
replace BinaryConnect with SBD-FQbinary weight AlexNet · ImageNet-2012 val, AlexNettop-1 accuracy (%): 35.4 -> 58.5Improved
quantize left/right singular vectors of SVD (Binary-SVD) instead of semi-binary alternating optimization (ablation)semi-binary components for AlexNet · ImageNet-2012 val, AlexNet, no fine-tuningtop-5 accuracy (%)Hurt
learn components with SBD-FQ instead of SBD-Direct when many layers are quantized (ablation)AlexNet without fine-tuning · ImageNet-2012 val, AlexNet, no fine-tuningtop-5 accuracy (%)Improved
increase beta (smaller K) in SBD-Direct (ablation)AlexNet without fine-tuning · ImageNet-2012 val, AlexNet, no fine-tuningtop-5 accuracy (%)Hurt
increase beta for fully-connected layers (ablation)AlexNet fc layers · ImageNet-2012 val, AlexNet, no fine-tuningtop-5 accuracy (%)No clear effect
deploy binary-weight AlexNet with 8-bit activations on FPGA instead of 32-bit floating pointFPGA accelerator · FPGA Xilinx Virtex-7 VX485Tlatency (ms): 21.6 -> 2.46Improved

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