AI Paper Atlas

Research map / Architectures and efficiency / Quantization

Quantized Neural Networks: Training Neural Networks with Low Precision Weights and Activations

Itay Hubara, Matthieu Courbariaux, Daniel Soudry, Ran El-Yaniv, Yoshua Bengio

JMLR 2018 · Architectures and efficiency › Quantization · approach: Mixed-precision quantization · 1,670 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
widen the hard tanh boundaries after training and retrain (ablation)BNN on AlexNet · ImageNettop-1 accuracy: improve top-1 accuracies by 1.5%Improved
use 2-bit activations instead of 1-bit with binary weightsBNN on AlexNet · ImageNettop-1 accuracy: 41.8% -> 51.03%Improved
use 4-bit weights and activations (QNN) instead of binaryGoogleNet · ImageNettop-1 accuracy: 47.1% -> 66.5%Improved
also quantize gradients to 6 bits (QNN 6-bit)GoogleNet QNN · ImageNettop-1 accuracy: 66.5% -> 66.4%No clear effect
replace vanilla BN and Adam with shift-based versionsBNN · MNIST/CIFAR-10accuracyNo clear effect
binarize RNN weights and activationsRNN/LSTM on Penn Treebank · Penn TreebankaccuracyHurt
increase RNN to 4-bit weights and activationsLSTM 300 units on Penn Treebank · Penn Treebankperplexity per word: 220 -> 100Improved
use XNOR GPU kernel instead of baseline kernelbinary MLP on MNIST · MNISTruntime: 7 times fasterImproved
use XNOR kernel for 8192x8192x8192 binary matmulGPU kernel · GTX750 matmulspeedup: 23 times faster than baseline, 3.4 times faster than cuBLASImproved

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