Explainable machine learning (2022, 5th lecture): Interpret. by backward prop. + concept discovery