主成分分析

From Ufldl

Jump to: navigation, search
Line 64: Line 64:
[[File:PCA-rotated.png|600px]]
[[File:PCA-rotated.png|600px]]
-
 
+
这就是把训练数据集旋转到 <math>\textstyle u_1</math><math>\textstyle u_2</math> 基后的结果。一般而言,运算 <math>\textstyle U^Tx</math> 表示旋转到基 <math>\textstyle u_1</math>,<math>\textstyle u_2</math>, ...,<math>\textstyle u_n</math> 之上的训练数据。矩阵 <math>\textstyle U</math> 有正交性,即满足  <math>\textstyle U^TU = UU^T = I</math> ,所以若想将旋转后的向量 <math>\textstyle x_{\rm rot}</math> 还原为原始数据 <math>\textstyle x</math> ,将其左乘矩阵<math>\textstyle U</math>即可: <math>\textstyle x=Ux_{\rm rot}</math> , 验算一下:
-
【原文】:This is the training set rotated into the <math>\textstyle u_1</math>,<math>\textstyle u_2</math> basis. In the general
+
-
case, <math>\textstyle U^Tx</math> will be the training set rotated into the basis
+
-
<math>\textstyle u_1</math>,<math>\textstyle u_2</math>, ...,<math>\textstyle u_n</math>.
+
-
 
+
-
One of the properties of <math>\textstyle U</math> is that it is an "orthogonal" matrix, which means
+
-
that it satisfies <math>\textstyle U^TU = UU^T = I</math>.
+
-
So if you ever need to go from the rotated vectors <math>\textstyle x_{\rm rot}</math> back to the
+
-
original data <math>\textstyle x</math>, you can compute
+
:<math>\begin{align}
:<math>\begin{align}
Line 79: Line 71:
because <math>\textstyle U x_{\rm rot} =  UU^T x = x</math>.
because <math>\textstyle U x_{\rm rot} =  UU^T x = x</math>.
-
【初译】:这就是训练数据集旋转到以<math>\textstyle u_1</math>,<math>\textstyle u_2</math>为基后的数据。但是通常情况下,训练数据集合通过<math>\textstyle U^Tx</math>旋转后会得到的是<math>\textstyle n</math>维的<math>\textstyle u_1</math>,<math>\textstyle u_2</math>, ...,<math>\textstyle u_n</math>。由于矩阵<math>\textstyle U</math>具有正交性,也就是说它满足<math>\textstyle U^TU = UU^T = I</math>,所以假若想从<math>\textstyle x_{\rm rot}</math>中还原<math>\textstyle x</math>,可以通过如下计算:
+
== 数据降维 ==
-
 
+
-
:<math>\begin{align}
+
-
x = U x_{\rm rot}  ,
+
-
\end{align}</math>
+
-
 
+
-
因为<math>\textstyle U x_{\rm rot} =  UU^T x = x</math>
+
-
 
+
-
【一审】:这就是训练数据集旋转到以<math>\textstyle u_1</math>,<math>\textstyle u_2</math>为基后的数据。但是通常情况下,训练数据集合通过<math>\textstyle U^Tx</math>旋转后会得到的是<math>\textstyle n</math>维的<math>\textstyle u_1</math>,<math>\textstyle u_2</math>, ...,<math>\textstyle u_n</math>。由于矩阵<math>\textstyle U</math>具有正交性,也就是说它满足<math>\textstyle U^TU = UU^T = I</math>,所以假若想从<math>\textstyle x_{\rm rot}</math>中还原<math>\textstyle x</math>,可以通过如下计算:
+
-
 
+
-
:<math>\begin{align}
+
-
x = U x_{\rm rot}  ,
+
-
\end{align}</math>
+
-
 
+
-
因为<math>\textstyle U x_{\rm rot} =  UU^T x = x</math>
+
-
 
+
-
 
+
-
【二审】:这就是训练数据集旋转到<math>\textstyle u_1</math>,<math>\textstyle u_2</math>基上所得数据。在更一般的情况下,训练数据集<math>\textstyle U^Tx</math>旋转到基<math>\textstyle u_1</math>,<math>\textstyle u_2</math>, ...,<math>\textstyle u_n</math>之上。矩阵<math>\textstyle U</math>具备正交性,也就是说它满足<math>\textstyle U^TU = UU^T = I</math>,所以假若想将旋转矩阵<math>\textstyle x_{\rm rot}</math>还原为原始数据<math>\textstyle x</math>,可以通过如下计算:
+
-
 
+
-
:<math>\begin{align}
+
-
x = U x_{\rm rot}  ,
+
-
\end{align}</math>
+
-
 
+
-
因为<math>\textstyle U x_{\rm rot} =  UU^T x = x</math>
+
-
 
+
-
 
+
-
 
+
-
== Reducing the Data Dimension 数据降维 ==
+

Revision as of 15:58, 18 March 2013

Personal tools