Better sampling in explanation methods can prevent dieselgate-like deception

Domen Vreš, Marko Robnik Šikonja

Machine learning models are used in many sensitive areas where besides predictive accuracy their comprehensibility is also important. Interpretability of prediction models is neces…