The Restricted Boltzmann Machine (RBM) is another technique that, composed of linear functions (which are usually called hidden units or neurons), creates a nonlinear transformation of the input data. The hidden units represent the status of the system, and the output dataset is actually the status of that layer.
The main hypothesis of this technique is that the input dataset is composed of features that represent probability (binary values or real values in the [0,1] range), since RBM is a probabilistic approach. In the following example, we will feed the RBM using binarized pixels of images as features (1=white, 0=black), and we will print the latent components of the system. These components represent different ...
