Redundancy in Feature Extraction
暂无分享,去创建一个
Given two random variables X and Y, a definition is offered that gives a condition for Y to be redundant with respect to X. It is shown that if such redundancy exists, then observations on Y, i.e., pattern vector elements related to Y, can be eliminated without increasing the classification error. A test for redundancy is developed and applied to the problem of preprocessing pattern vectors to eliminate redundant vector elements.