An information-theoretic learning model based on importance sampling

Jiangshe Zhang, Lizhen Ji, Fei Gao +1

A crucial assumption underlying the most current theory of machine learning is that the training distribution is identical to the test distribution. However, this assumption may no…