Towards Robust And Fair Machine Learning