Towards Trustworthy Machine Learning