Skip to content
Implementation for the paper "Latent Weights Do Not Exist: Rethinking Binarized Neural Network Optimization"
Branch: master
Clone or download
koenhelwegen Merge pull request #22 from plumerai/poster-margins
Remove print padding from poster
Latest commit 9856919 Dec 8, 2019
Type Name Latest commit message Commit time
Failed to load latest commit information.
bnn_optimization Update to camera ready version (#18) Oct 31, 2019
.gitignore Bop hyperparameter analysis (#2) Jun 3, 2019
poster.pdf Remove print padding Dec 8, 2019
requirements.txt Update to camera ready version (#18) Oct 31, 2019

Rethinking Binarized Neural Network Optimization

arXiv:1906.02107 License: Apache 2.0 Code style: black

Implementation for paper "Latent Weights Do Not Exist: Rethinking Binarized Neural Network Optimization".

A poster illustrating the proposed algorithm and its relation to the previous BNN optimization strategy is included at ./poster.pdf.

Note: Bop is now added to Larq, the open source training library for BNNs. We recommend using the Larq implementation of Bop: it is compatible with more versions of TensorFlow and will be more actively maintained.


You can also check out one of our prebuilt docker images.


This is a complete Python module. To install it in your local Python environment, cd into the folder containing and run:

pip install -e .


To train a model locally, you can use the cli:

bnno train binarynet --dataset cifar10

Reproduce Paper Experiments

Hyperparameter Analysis (section 5.1)

To reproduce the runs exploring various hyperparameters, run:

bnno train binarynet \
    --dataset cifar10 \
    --preprocess-fn resize_and_flip \
    --hparams-set bop \
    --hparams threshold=1e-6,gamma=1e-3

where you use the appropriate values for threshold and gamma.

CIFAR-10 (section 5.2)

To achieve the accuracy in the paper of 91.3%, run:

bnno train binarynet \
    --dataset cifar10 \
    --preprocess-fn resize_and_flip \
    --hparams-set bop_sec52 \

ImageNet (section 5.3)

To reproduce the reported results on ImageNet, run:

bnno train alexnet --dataset imagenet2012 --hparams-set bop
bnno train xnornet --dataset imagenet2012 --hparams-set bop
bnno train birealnet --dataset imagenet2012 --hparams-set bop

This should give the results listed below. Click on the tensorboard icons to see training and validation accuracy curves of the reported runs.

Network Bop - top-1 accuracy
Binary Alexnet 41.1% tensorboard
XNOR-Net 45.9% tensorboard
Bi-Real Net 56.6% tensorboard
You can’t perform that action at this time.