首页> 外国专利> SETTING DEVICE, INFORMATION CLASSIFICATION DEVICE, CLASSIFICATION PLANE SETTING METHOD OF SETTING DEVICE, AND INFORMATION CLASSIFICATION METHOD AND PROGRAM OF INFORMATION CLASSIFICATION DEVICE

SETTING DEVICE, INFORMATION CLASSIFICATION DEVICE, CLASSIFICATION PLANE SETTING METHOD OF SETTING DEVICE, AND INFORMATION CLASSIFICATION METHOD AND PROGRAM OF INFORMATION CLASSIFICATION DEVICE

机译:设置设备,信息分类设备,设置设备的分类平面设置方法以及信息分类设备的信息分类方法和程序

摘要

PROBLEM TO BE SOLVED: To reduce the processing load of exponentiation calculation of kernel even when the number of elements of a feature vector increases.SOLUTION: Feature vectors are extracted from information for learning, and a classification plane allowing correct classification of the information for learning indicating how to classify is formed. As the kernel used for a nonlinear support vector machine used for forming the classification plane, a function expression of exponentiating the calculation value of the calculation of determining the similarity between the feature vectors is used. A plurality of classification planes are formed while the combination of a regularization parameter indicating the tolerance of false classification and an exponentiation parameter of the kernel is changed. Using the plurality of formed classification planes, information for a test indicating how to classify is classified, and a classification plane allowing the information for a test to be most accurately classified is selected and is set in an information classification device.SELECTED DRAWING: Figure 7
机译:要解决的问题:即使特征向量的元素数量增加,也要减少内核的幂运算处理量解决方案:从学习信息中提取特征向量,并通过分类平面对学习信息进行正确分类指示如何分类。作为用于用于形成分类平面的非线性支持向量机的内核,使用对确定特征向量之间的相似性的计算的计算值取幂的函数表达式。形成多个分类平面,同时改变指示错误分类的容忍度的正则化参数和内核的幂参数的组合。使用形成的多个分类平面,对用于指示如何分类的测试信息进行分类,并且选择允许最准确地对测试信息进行分类的分类平面,并将其设置在信息分类设备中。所选择的图:图7

著录项

  • 公开/公告号JP2016095677A

    专利类型

  • 公开/公告日2016-05-26

    原文格式PDF

  • 申请/专利权人 CASIO COMPUT CO LTD;

    申请/专利号JP20140231395

  • 发明设计人 MATSUNAGA KAZUHISA;

    申请日2014-11-14

  • 分类号G06N99;G06F17/30;

  • 国家 JP

  • 入库时间 2022-08-21 14:46:16

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号