https://github.com/uranusx86/binarynet-on-tensorflow

binary weight neural network implementation on tensorflow
https://github.com/uranusx86/binarynet-on-tensorflow

deep-learning neural-network python tensorflow tensorflow-experiments

Last synced: 3 days ago
JSON representation

binary weight neural network implementation on tensorflow

• Host: GitHub
• URL: https://github.com/uranusx86/binarynet-on-tensorflow
• Owner: uranusx86
• Created: 2018-02-12T12:58:40.000Z (almost 7 years ago)
• Default Branch: master
• Last Pushed: 2018-03-23T13:14:28.000Z (over 6 years ago)
• Last Synced: 2024-11-12T17:03:58.490Z (3 days ago)
• Topics: deep-learning, neural-network, python, tensorflow, tensorflow-experiments
• Language: Python
• Size: 10.8 MB
• Stars: 54
• Watchers: 3
• Forks: 24
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# BinaryNet-on-tensorflow

binary weight neural network implementation on tensorflow

This is an implementation code for reproducing BNN

## How to run

```

python mnist.py

python cifar10.py

```

## Accuracy

| DataSet | accuracy |

|---------|----------|

| MNIST   |  99.04%  |

| Cifar10 |  86.18%  |

## Different between paper

layer-wise learning rate, paper is layer_lr = 1./sqrt(1.5 / (num_inputs + num_units)), my implement is layer_lr / 4

## Ref

```

[BinaryNet: Training Deep Neural Networks with Weights and Activations Constrained to +1 or -1"](http://arxiv.org/abs/1602.02830),

Matthieu Courbariaux, Yoshua Bengio

```

and

[An implemtation of binaryNet for Keras](https://github.com/DingKe/nn_playground/tree/master/binarynet)