Currently, the Balios GUI supports only parametric learning, i.e., the process of computing the parameters best fitting a given set of data cases. More precisely, Balios supports EM (Expectation-Maximization) and gradient-based optimization methods..
Reconsider the blood type Bayesian logic program from the Logical Predicates Section, i.e., we we have used logical background knowledge to describe the founders of families. In order to estimate the parameters, we need some data cases, i.e., (partially observed) Herbrand models of Bayesian logic program. As we do not have such data yet, let us simply generate it. To so so, pose a query, such as P(bloodtype(kristian)), and press the sample button in the tool bar in the Supportnetwork pane.
Pressing the Create Data Cases will sample 1000 data cases into the file you specified, say blood_founder.data (<- Please have a look at the file to see the data format).
Now that we have a data case, we can re-estimate the paramters od the BLP. Press the EM button Select the Browse button and choose a file from which the associated conditional probability distribution must be computed; in our case blood_founder.data. Choose the parameter estimation method, specify all parameters, and hit the Learn. After a while, your can browse the BLP file to see the estimated distributions.