Naming of models

[karalets]

At the danger of being a Bayesian fascist here, let me make some recommendations for naming conventions.

I.e. the file gaussian_variational.py is just defining a maximum likelihood graph_nn, the word variational is not just misleading here but there is nothing variational happening at all, as this is an inference property.

Suggestion: Name models for model properties, i.e. model_graph_nn_gaussian.py or so. And we can name inference things accordingly.

[yuanqing-wang]

fixed.

but we might get rid of this script to further separate things eventually