Removing -dependence in minimal interpretable model learning: distribution conditions and structural parameters | Synapse