Cyan's Blog

Search

Search IconIcon to open search

MIT_18.065-Part_6-Orthonormal Columns in Q Give Q'Q = I

Last updated Nov 13, 2021 Edit Source

# Orthonormal Columns in Q Give Q’Q = I

2021-11-13

Tags: #Math/LinearAlgebra

orthogonal: 正交的 orthonormal: Ortho(gonal) + normal, 即又正交又是单位向量, 长度为1.

# Orthogonal Vectors

内积的向量化表示: $\boldsymbol{x}^{\mathrm{T}} \boldsymbol{y}=x_{1} y_{1}+\cdots+x_{n} y_{n}=0$

需要注意的是, 如果是复数相连的内积, 我们需要使用共轭复数(Conjugate)组成的向量来计算内积: $\overline{\boldsymbol{x}}^{\mathrm{T}} \boldsymbol{y}=\bar{x}{1} y{1}+\cdots+\bar{x}{n} y{n}=0$

# 与勾股定理的联系

$$\begin{aligned}||x-y||^2&=(x-y)^T(x-y)\&=x^Tx+y^Ty-x^Ty-y^Tx\&=||x||^2+||y||^2-2||x||\space ||y||cos\theta\end{aligned}$$

# Orthogonal Matrix

有了向量相互正交的表示, 我们可以很容易的推广到矩阵:

$$Orthogonal\space matrix: Q^{\mathrm{T}} Q=QQ^{\mathrm{T}}=I_{n\times n}$$

# 重要矩阵列举

# Rotation Matrix & Reflection Matrix

$$Q_{\text {rotate }}=\left[\begin{array}{rr} \cos \theta & -\sin \theta \\ \sin \theta & \cos \theta \end{array}\right]=\text { rotation through an angle } \theta$$

$$Q_{\text {reflect }}=\left[\begin{array}{rr} \cos \theta & \sin \theta \\ \sin \theta & -\cos \theta \end{array}\right]=\text { reflection across the } \frac{\theta}{2}\text { line. }$$

# Hadamard matrices

# Householder Reflections

两个例子: $$\boldsymbol{H}{3}=I-\frac{2}{3} \text { ones }=\frac{1}{3}\left[\begin{array}{rrr} 1 & -2 & -2 \\ -2 & 1 & -2 \\ -2 & -2 & 1 \end{array}\right] \quad \boldsymbol{H}{4}=I-\frac{2}{4} \text { ones }=\frac{1}{2}\left[\begin{array}{rrrr} \mathbf{1} & -1 & -1 & -1 \\ -1 & \mathbf{1} & -1 & -1 \\ -1 & -1 & \mathbf{1} & -1 \\ -1 & -1 & -1 & \mathbf{1} \end{array}\right]$$

# Haar wavelets

n=8 $$\left[\begin{array}{ccc}1&1&1&&1&&&\\ 1&1&1&&-1&&&\\ 1&1&-1&&&1&&\\ 1&1&-1&&&-1&&\\ 1&-1&&1&&&1&\\ 1&-1&&1&&&-1&\\ 1&-1&&-1&&&&1\\ 1&-1&&-1&&&&-1\end{array}\right]$$

# Eigenvectors of $S=S^T$ & $Q^TQ=I$

注意复数的内积需要取共轭: 验证, 对于F, 第二列和第四列的内积: $[1,-i,i^2,-i^3]\left[\begin{array}{ccrr}1 \i^3 \i^6 \\ i^9\end{array}\right]=1-1+1-1=0$

# 每一个向量空间$R^n$都有一组正交基

$$\text { Singular vectors } \quad A \boldsymbol{v}{1}=\sigma{1} \boldsymbol{u}{1} \quad A \boldsymbol{v}{2}=\sigma_{2} \boldsymbol{u}{2} \quad \cdots \quad A \boldsymbol{v}{r}=\sigma_{r} \boldsymbol{u}_{r}$$

# 投影矩阵

$$Px=QQ^{\mathrm{T}}x=Q(Q^{\mathrm{T}}x)=$$ $$Col(Q)(Inner\space product\space of\space Row(Q)\space and\space x)$$ 就是先计算x与Q里面各个正交基底的内积, 得到在这个正交基底下的"坐标", 然后再用Col(Q)表示出来.