Examples Should not require a dataset which is not in the repo

We should use some sklearn functions to create/load test datasets.

Fact examples should be moved in a branch. In this branch can be updated from the master and contain all specific FACT examples.