Let's now create our first RBM in scikit-learn. We will start by instantiating a module to extract 100 components from our MNIST dataset.

We will also set the verbose parameter to True to allow us visibility into the training process as well as the random_state parameter to 0. The random_state parameter is an integer that allows for reproducibility in code. It fixes the random number generator and sets the weights and biases randomly at the same time, every time. We finally let n_iter be 20. This is the number of iterations we wish to do, or back and forth passes of the network:

# instantiate our BernoulliRBM
 # we set a random_state to initialize our weights and biases to the same starting point
 # verbose is set to True to see the fitting period
 # n_iter is the number of back and forth passes
 # n_components (like PCA and LDA) represent the number of features to create
 # n_components can be any integer, less than , equal to, or greater than the original...