《Discriminative learning forminimum error classification》.pdf

《Discriminative learning forminimum error classification》.pdf

  1. 1、本文档共12页,可阅读全部内容。
  2. 2、原创力文档(book118)网站文档一经付费(服务费),不意味着购买了该文档的版权,仅供个人/单位学习、研究之用,不得用于商业用途,未经授权,严禁复制、发行、汇编、翻译或者网络传播等,侵权必究。
  3. 3、本站所有内容均由合作方或网友上传,本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺!文档内容仅供研究参考,付费前请自行鉴别。如您付费,意味着您自己接受本站规则且自行承担风险,本站不退款、不进行额外附加服务;查看《如何避免下载的几个坑》。如果您已付费下载过本站文档,您可以点击 这里二次下载
  4. 4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等,请点击“版权申诉”(推荐),也可以打举报电话:400-050-0827(电话支持时间:9:00-18:30)。
查看更多
IEEE TRANSACTIONS ON SIGNAL PROCESSING. VOL. 40, NO. 12, DECEMBER 1992 3043 Discriminative Learning for Minimum Error Classification Biing-Hwang Juang, Fellow, IEEE, and Shigeru Katagiri, Member, IEEE Abstract-Recently, due to the advent of artificial neural net- is a penalty or cost associated with a misclassification, the works and learning vector quantizers, there is a resurgent in- objective is then to minimize the expected cost accord- terest in reexamining the classical techniques of discriminant ingly. analysis to suit the new classifier structures. One of the partic- ular problems of interest is minimum error classification in Bayes decision theory [l] is a fundamental statistical which the misclassificationprobability is to be minimized based approach to the classification problem and is often the ba- on a given set of training samples. In this paper, we propose a sis of many pattern classification techniques. Suppose we new formulation for the minimum error classification problem, have full knowledge of the a posteriori probability together with a fundamental technique for designing a classifier PA(C;Ix), which is defined by the parameter set A as de- that approaches the objective of minimum classification error in a more direct manner than traditional methods. We contrast noted by the subscript. (Note that this assumption also the new method to several traditional classifier designs in typ- implies that the true a posteriori probability can be para- ical experiment

文档评论(0)

ghfa + 关注
实名认证
内容提供者

该用户很懒,什么也没介绍

1亿VIP精品文档

相关文档