Some ways to intepretate black-box models