Standard machine learning techniques assume that the statistical structure of the training and test datasets are the same (i.e. same attribute distribution p(x), and same class distribution p(c|x)). However, in real ...