非参数估计(第1讲)详解.ppt

* * 还是一个遍历方法，只是到某些子根节点就不必走下去，就可以返回了。 * 流程图 * 红黑两类，绿圈为类中心区域 * C-E * C-E * 分类 理由 产生分类器 窗宽度是一个平滑参数 * 窗宽度是一个平滑参数 * * How these limitations can be circumvented if an unlimited number of samples is available. * 寻找V的一个折中 * * 对于任意j=1，，，D 朴素估计器 * * 用某种核函数表示某一样本对待估计的密度函数的贡献，所有样本所作贡献的线性组合视作对某点概率密度p(x)的估计 * ??? Pn * ????? * **** * 前面我们看到的是一个单峰的经典密度分布样本估计概率密度的结果,下面我们再看看多峰的分布样本估计情况 待估的密度函数是这样的 还是使用正态窗 * generalize * ?In high dimensional space, randomly generated points are approximately equidistant - almost all points classified the same way？？？ * ？？？ * C-e * E-c * * 同K近邻估计两个一维密度：高斯和双峰分布。请注意，用有限的N个样本估计,结果是相当“高低不平” * 同K近邻估计两个一维密度：高斯和双峰分布。请注意，用有限的N个样本估计,结果是相当“高低不平” * E-c * * 划分特征空间为小的单元 边界线在两点之间的中间线处 用单元内包裹的点的类型来标示单元 如图有红黑两类 * E-C * E-C * E-C * E-C * Example of k-NN rule with k = 3 There are two classes: yellow triangles and blue squares. The circle represents the unknown sample x and as two of its nearest neighbors come from class ω2, it is labeled class ω2. The number k should be: large to minimize probability of misclassifying x. small (with respect to no of samples) so that points are close enough to x to give an accurate estimate of the true class of x. Nearest Neighbor and Voronoi Tesselation最近邻规则和Voronoi 网格 Voronoi Tesselation：Decision surface formed by the training examples 决策面来自于训练样本 预备知识：Voronoi图，沃罗诺伊图，又叫泰森多边图或Dirichlet图，它是由一组连接两邻近点直线的垂直平分线组成的连续多边形组成。由此划分的每个区域中包含一个点，对于点pi，它所在区域中的任何一个位置距离pi点比距离其他点更近。 Voronoi Diagrams Decision boundary is a polygon such that any point that falls in is closer to than any other sample. Vi是一个多边形,任何落入这个多边形的点x距离点Si都比其他样本点的距离要近。 Nearest Neighbor and Voronoi Tesselation最近邻规则和Voronoi 网格 Voronoi Tesselation：Decision surface formed by the training examples 决策面来自于训练样本 Nearest Neighbor and Voronoi Tesselation最近邻规则和Voronoi 网格 Partition the feature space into cells. Boundary lines lie half-way between any two points. Label each cell based on the class of enclosed point. 2 classes: red, black