fisher准则线性分类器matlab,fisher准则线性分类器设计.doc

fisher准则线性分类器设计.doc

一、基于FISHER准则线性分类器设计1、实验内容已知有两类数据和二者的概率已知1P06，04。122P中数据点的坐标对应一一如下1数据X023311520706499077571052411974029080251806682056220902301333054310940702126005070081007315033451065000247010430312206655058381165312653081370339905152072260201504070017171057302099Y233852194616730163651784420155206812121324797151181969218340187042294817714239391564819329220272456817523169912488317259204662022623757179872082820798194492380122373216141923522604Z053380851410831041641117605536060710443904928059011092710756100720427204353098690484110992102990712710124045760854411275077050412910085076760841808784097510784004158103150753309548数据点的对应的三维坐标为2X2140101230120814116551374011829176321973924152258902847219539125001286412614200712183117909133221146617087159202935314664293131834918340250962719823148203532603012327214651567329414Y2102980961109154149010820009399114051067808050128891460114334070911294213744093871226611833087980559205150099830912007126128331102912680071401244613392118080550314708114350767911288Z2062101365605498067080893214342095080732405784149431091507644121591304911408093980619706603139281408406909084000538113729077310731913439081420958607379075480739306739086511369911458数据的样本点分布如下图2101230511522500511521请把数据作为样本，根据FISHER选择投影方向的原则，使原样本W向量在该方向上的投影能兼顾类间分布尽可能分开，类内样本投影尽可能密集的要求，求出评价投影方向的函数，并在图形表示出来。并在实验报告中表示出来，并求使取极大值的。WJF用MATLAB完成FISHER线性分类器的设计，程序的语句要求有注释。2根据上述的结果并判断(1，15，06)12，10，055，20，09，068，12，15，089，(023，233，143)，属于哪个类别，并画出数据分类相应的结果图，要求画出其在上的投影。W3回答如下问题，分析一下的比例因子对于FISHER判别函数没有影W响的原因。2、实验代码X1023311520706499077571052411974029080251806682056220902301333054310940702126005070081007315033451065000247010430312206655058381165312653081370339905152072260201504070017171057302099X2233852194616730163651784420155206812121324797151181969218340187042294817714239391564819329220272456817523169912488317259204662022623757179872082820798194492380122373216141923522604X3053380851410831041641117605536060710443904928059011092710756100720427204353098690484110992102990712710124045760854411275077050412910085076760841808784097510784004158103150753309548将X1、X2、X3变为行向量X1X1X2X2X3X3计算第一类的样本均值向量M1M11MEANX1M12MEANX2M13MEANX3计算第一类样本类内离散度矩阵S1S1ZEROS3,3FORI136S1S1M11X1IM12X2IM13X3I M11X1IM12X2IM13X3IENDW2的数据点坐标X4140101230120814116551374011829176321973924152258902847219539125001286412614200712183117909133221146617087159202935314664293131834918340250962719823148203532603012327214651567329414X5102980961109154149010820009399114051067808050128891460114334070911294213744093871226611833087980559205150099830912007126128331102912680071401244613392118080550314708114350767911288X6062101365605498067080893214342095080732405784149431091507644121591304911408093980619706603139281408406909084000538113729077310731913439081420958607379075480739306739086511369911458X4X4X5X5X6X6计算第二类的样本均值向量M2M21MEANX4M22MEANX5M23MEANX6计算第二类样本类内离散度矩阵S2S2ZEROS3,3FORI136S2S2M21X4IM22X5IM23X6I M21X4IM22X5IM23X6IEND总类内离散度矩阵SWSWZEROS3,3SWS1S2样本类间离散度矩阵SBSBZEROS3,3SBM1M2 M1M2最优解WWSW1M1M2 将W变为单位向量以方便计算投影WW/SQRTSUMW2计算一维Y空间中的各类样本均值M1及M2FORI136YIW