This class contains examples about how we can create CajaMar's dynamic models using the AMIDST Toolbox.
It show how to create 2T-DBNs over multinomial, Gassuian and IndicatorDistribution variables.
Created by andresmasegosa on 22/11/14.
In this example, we create the proposed dynamic model for making predictions about the defaulting
behaviour of a client. We took some fake data with some fake attributes.
We show how to create indicator variables and use it in the model.
We finally compute the log-likelihood of the data according to the created model (i.e. the probabilty distributions
are randomly initialized, there is no parametric learning). The data is a single long temporal sequence.