ML interpretability fundamentals