In the previous example on the bank marketing dataset, we observed about 89% classification accuracy using MLP. We also normalized the original dataset before feeding it to the MLP.
In this section, we will see how to use the same datasets for the DBN-based predictive model. We will use the customized and extended version of DBN implantation called deep-belief-network that can be downloaded from GitHub at https://github.com/albertbup/deep-belief-network. The deep-belief-network is a simple, clean, fast Python implementation of deep belief networks based on binary Restricted Boltzmann Machines (RBM), built upon NumPy and TensorFlow libraries in order to take advantage of GPU computation. This library is implemented based on the following two research papers:
Hinton, Geoffrey E., Simon Osindero, and Yee-Whye Teh. A fast learning algorithm for deep belief nets Neural Computation 18.7 (2006): 1527-1554.
Fischer, Asja, and Christian Igel. Training...