[模式识别第6章.ppt

第6章 近邻法Chapter 6: Nearest Neighbor method Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. 6.1 近邻法的基本原理 6.2 k-近邻法的基本原理 6.3 近邻法的分类错误率分析 6.4 近邻法的改进算法 6.5 最佳距离度量近邻法 本章主要内容 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. 6.1 近邻法的基本原理 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. 6.1.1 关于近邻法 近邻法原理示意图 1951年Fix和Hodges首次提出 最经典的模式识别方法之一 方法简单，便于理论分析 是其它模式识别方法的标尺 “距离”的度量方式有很多种 第 类 第 类 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. 近邻法原理及其决策规则 最小距离分类器 它将各类训练样本划分成若干子类 并在每个子类中确定代表点 用子类的质心或邻近质心的某一样本为代表点 测试样本的类别则以其与这些代表点距离最近作决策 缺点: 所选择的代表点并不一定能很好地代表各类，其后果将使错误率增加 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. 最小距离分类器 每个类别只有一个”代表点” Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. 基于距离的分段线性函数 每个类别用多个”代表点”表示 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. Evaluation only. Created with