Skip to content

Latest commit

 

History

History
37 lines (33 loc) · 1.17 KB

README.md

File metadata and controls

37 lines (33 loc) · 1.17 KB
Raw

Introduction

rbm.py is the fastest and easiest way to use Restricted Boltzmann Machines (RBMs). RBMs are a class of probabilistic models that can discover hidden patterns in your data. rbm.py provides all the necessary methods with a pythonic interface, and moreover, all methods call blazing fast C code. The code can also run transparently on GPU thanks to Theano (http://deeplearning.net/software/theano/).

Here's an example usage

$ python
>>> import numpy, rbm
>>> X = numpy.array([[0, 1, 0], \
                     [0, 1, 1], \
                     [1, 0, 1], \
                     [1, 1, 1]]) # Improvised dataset
>>> model = rbm.RBM(n_hiddens=2) # RBM with two hiddens units
>>> model.fit(X) # Train using dataset X
>>> model.sample_h(X) # Get hidden code
array([[0, 0],
       [0, 1],
       [1, 1],
       [1, 0]])
>>> model.gibbs(X) # MCMC step
array([[0, 1, 1],
       [1, 0, 1],
       [0, 0, 1],
       [1, 0, 1]])

Authors and Contributors

Yann N. Dauphin (@ynd)

Support or Contact

Having trouble? Check out https://github.com/ynd/rbm.py/issues.