https://github.com/jasonwbw/effictiverbm

python version restricted boltzmann machine tools for solve problems or build more complex model
https://github.com/jasonwbw/effictiverbm

Last synced: 19 days ago
JSON representation

python version restricted boltzmann machine tools for solve problems or build more complex model

• Host: GitHub
• URL: https://github.com/jasonwbw/effictiverbm
• Owner: jasonwbw
• License: mit
• Created: 2014-05-19T15:49:15.000Z (over 10 years ago)
• Default Branch: master
• Last Pushed: 2014-08-27T06:34:16.000Z (about 10 years ago)
• Last Synced: 2023-10-20T20:39:30.811Z (about 1 year ago)
• Language: Python
• Size: 11.6 MB
• Stars: 1
• Watchers: 3
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
EffictiveRBM

============

EffictiveRBM is a python version restricted boltzmann machine tools for solve problems or build more complex model. The effictive means this is a fast rbm tools, it accelerate by cython and multi-process. It has been used to build the autoencoder model which descript by Hinton.

## How to Use

First, initialize an RBM with the desired number of visible, hidden units and process number.  

Then train the rbm model with Iterator and params which descript in rbm.py

```python

rbm = ParallelRBM(self.converter.dimensionality, 1000, 5)

rbm.train(DefaultBatchIterator(100, self.converter.train_images), max_epochs = 20, batch = 100)

```

## Dependencies

- [NumPy](http://www.numpy.org): normal computing

- [Scipy](http://www.scipy.org/): faster sigmoid computing

## Optional dependencies

- [Cython](http://www.cython.org): Only necessary to run the faster version. Version 0.17.1 or higher.

- [matplotlib](http://matplotlib.sourceforge.net/): For plotting some data

## Reference

* [A Practical guide to training restricted Boltzmann machines](http://www.cs.toronto.edu/~hinton/absps/guideTR.pdf), by Geoffrey Hinton.

* [Reducing the Dimensionality of data with neural networks](http://www.cs.toronto.edu/~hinton/science.pdf), by Geoffrey Hinton.

* [Learning Deep Architectures for AI](http://www.iro.umontreal.ca/~lisa/pointeurs/TR1312.pdf), by Yoshua Bengio.

## License

The MIT License (MIT)