Local information embedding by dimension extension and information risk minimization
Qun Chang, Xuan hang Wang, Xiao-Long Wang · 2005
Information risks such as information misunderstanding and insufficiencies of metric will affect learning quality. Information risk minimization (IRM) is proposed aiming to solve this problem. Besides, towards IRM, the framework of dimension extension is introduced to make it possible that more local, additional, or prior information can be added into learning course.