Can we trust AI? towards practical implementation and theoretical analysis in trustworthy machine learning