Restricted-Boltzmann-Machine

Python implementation of Restricted Boltzmann Machine (RBM). And an example on MNIST dataset. Implementation code is in RBM.py and it's use for MNIST data is in the notebook rbm_mnist_example.ipynb.

RBM code is very well documented for using it for your data.

Requirements

• Numpy
• Scipy
• Matplotlib (for MNIST example)
• time (for MNIST example)
• struct (for MNIST example)

Results on MNIST

• 100 Hidden units

• Error plot for 20 epochs

  

• Reconstruction of few selected images as the model gets trained

  

• Learned weights of the RBM over time

  

Todo

• (Block) Gibbs sampling - take average instead of taking single value.
• Persistent Contrastive Divergence.
• Fast PCD.

References

[1] An Introduction to Restricted Boltzmann Machines - Asja Fischer and Christian Igel
[2] Training Restricted Boltzmann Machines: An Introduction - Asja Fischer and Christian Igel
[3] A Practical Guide to Training Restricted Boltzmann Machines - Geoffrey Hinton