Flexible Discriminant Analysis Penalized Discriminant Analysis Mixture Discriminant Analysi.pdfVIP

下载本文档

11
0
约7.03万字
约 28页
2017-04-13 发布于江苏
举报
版权申诉

Flexible Discriminant Analysis Penalized Discriminant Analysis Mixture Discriminant Analysi.pdf

1、本文档共28页，可阅读全部内容。
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、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、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)。
5、该文档为VIP文档，如果想要下载，成为VIP会员后，下载免费。
6、成为VIP后，下载本文档将扣除1次下载权益。下载后，不支持退款、换文档。如有疑问请联系我们。
7、成为VIP后，您将拥有八大权益，权益包括：VIP文档下载权益、阅读免打扰、文档格式转换、高级专利检索、专属身份标志、高级客服、多端互通、版权登记。
8、VIP文档为合作方或网友上传，每下载1次，网站将根据用户上传文档的质量评分、类型等，对文档贡献者给予高额补贴、流量扶持。如果你也想贡献VIP文档。上传文档

Flexible Discriminant Analysis Penalized Discriminant Analysis Mixture Discriminant Analysi

Support Vector Machines and Flexible Discriminants Chapter 12: Hastie et al. (2001) Madhusudana Shashanka Department of Cognitive and Neural Systems Boston University CN700 - SVMs/Flexible Discriminants March 23, 2004 – p.1/28 Overview Part I – SVMs Support Vector Classifier Support Vector Machines Part II – Flexible Discriminants Flexible Discriminant Analysis Penalized Discriminant Analysis Mixture Discriminant Analysis CN700 - SVMs/Flexible Discriminants March 23, 2004 – p.2/28 Introduction Generalizations of linear decision boundaries for classification. Optimal separating hyperplanes for linearly separable classes. Extensions to the non-separable case generalize to support vector machines. SVM - produce nonlinear decision boundaries by constructing a linear boundary in a large, transformed space. CN700 - SVMs/Flexible Discriminants March 23, 2004 – p.3/28 Separating Hyperplanes - Recap Consider classes separable by a linear boundary. Data of the form (xi, yi) with xi ∈ Rp, yi ∈ {?1, 1} ?i = 1, 2, . . . , N . Define a hyperplane {x : f(x) = xT β + β0 = 0}. Find the hyperplane that creates the biggest margin between the training points for classes -1 and 1. minβ,β0 1 2 ||β||2 subject to yi(xTi β + β0) ≥ 1, i = 1, 2, . . . , N . Margin is C = 1/||β||. Classification rule G(x) = sign[xT β + β0]. Convex optimization problem: quadratic criterion and linear inequality constraints. CN700 - SVMs/Flexible Discriminants March 23, 2004 – p.4/28 Non-separable classes margin A A F ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?? ca d b e A A F margin Allow some points on the wrong side. Define slack variables ξ = (ξ1, ξ2, . . . , ξn). Modify the constraint as: yi(xTi β + β0) ≥ C(1 ? ξi), ?i, ξi ≥ 0, ∑N i=1 ξi ≤ constant. Misclassifications occur when ξi 1. CN700 - SVMs/Flexible Discriminants March 23, 2004 – p.5/28 Support Vector Classifier minβ,β0 1 2 ||β||2 + γ ∑N i=1 ξi subject to ξi ≥ 0, yi(xTi βxi + b) ≥ 1 ? ξi?i. LP = 1 2 ||β||2 + γ