Numpy implementation of Restricted Boltzmann Machine.
This model generates MNIST images using RBM. Constrative Divergence(CD) and persistent CD(PCD) is implemeted.
Deep Belief Network can construct simply by stacking RBM. In DBN case, input dimension of later RBM layer should be equal to hidden dimension of forrmer RBM layer.
- main.py: Main function of implemenation, construct the model, generates images, and calculate entropy of hidden unit
- model.py: RBM class(CD, PCD algorithms)
- downlad.py: Files for downlading MNIST data sets
- ops.py: Operation functions
- utils.py: Functions dealing with images processing.
First, download dataset with:
$ python download.py mnist
Second, write the main function with configuration you want.
