Going through these examples is probably the fastest way to understand how to use Marbl. These are all extremely trivial problems, chosen so that the data formats for inputs and outputs can be shown in full detail.
This is explained in the single-node inference example
This is explained in the chain inference example
The first learning example consists of learning a model with a single output variable, which is equivalent to learning a standard (multivariate, non-structured) logistic regression model. This is explained in the logistic regression example.
There is also an example of how to do logistic regression using MPI
A simple chain CRF with four nodes is learned in chain learning.