Invited talk: On estimating epistemic uncertainty