MIT18.06 Linear Algebra (24)

lec24

Markov Matrix

A=[.1.01.3.2.99.3.70.4]A=\left[\begin{array}{ccc}.1 & .01 & .3 \\.2 & .99 & .3 \\.7 & 0 & .4\end{array}\right]

properties

  1. All entryis 0\ge 0
  2. All columns add to 11
  3. λ=1\lambda =1 is an eigen value
  4. all other eigen value λi<1|\lambda_i|<1

steady state

uk=Aku0=c1λ1x1+c2λ2x2++cnλnxnu_k=A^ku_0=c_1\lambda _1x_1+c_2\lambda _2x_2+\cdots+c_n\lambda _nx_n

if λ1=1\lambda_1=1 , the steady state of uku_k ( kk\to \infty ) is c1x1c_1x_1 i.e. the x1x_1 part of u0u_0 .

why λ=1\lambda =1 is an eigenvalue?

A1I=[.9.01.3.2.01.3.70.6]A-1 I=\left[\begin{array}{ccc}-.9 & .01 & .3 \\.2 & -.01 & .3 \\.7 & 0 & -.6\end{array}\right]

All columns in AIA-I adds up to 00 $\stackrel{?}{\longrightarrow} $ $ A-I$ is singular

Beacuse rows are dependent.

row1+row2++rown=0row_1+row_2+\cdots+row_n=0

i.e. (1,1,1)(1,1,1) is in N((AI))N((A-I)^\top)

Application of Marcov matrix: population

[ucalumass]0=[01000]\begin{bmatrix}u_{cal} \\u_{mass}\end{bmatrix}_{0}=\begin{bmatrix}0\\1000\end{bmatrix}

[ucalumass]k+1 =[.9.2.1.8][ucalumass]k\begin{bmatrix}u_{cal} \\u_{mass}\end{bmatrix}_{k+1} =\begin{bmatrix}.9 & .2 \\.1 & .8\end{bmatrix}\begin{bmatrix}u_{cal} \\u_{mass}\end{bmatrix}_{k}

$.1 ofCaliforniapopulationof California population \to$ Massachusetts

$.2 ofMassachusettspopulationof Massachusetts population \to$ California

find eigenvalue and eigenvectors:

λ=1,.7 \lambda = 1, .7

u0=[01000]=10003[21]+20003[11]u_0=\begin{bmatrix}0 \\ 1000\end{bmatrix}=\frac{1000}{3} \begin{bmatrix}2 \\ 1\end{bmatrix}+\frac{2000}{3} \begin{bmatrix}-1 \\ 1\end{bmatrix}

uk=c11k[21]+c2(.7)k[11]u_{k}=c_{1}{\color{Red} 1} ^{k}\begin{bmatrix} 2 \\ 1 \end{bmatrix}+c_{2}(.7)^{k}\begin{bmatrix} -1 \\ 1 \end{bmatrix}

steady state

uk=10003[21]u_k=\frac{1000}{3} \begin{bmatrix}2 \\ 1\end{bmatrix}

Fourier Series

intro

xix_i are orthonormal basis.

v=x1q1+x2q2++xnqnv=x_1q_1+x_2q_2+\dots+x_nq_n

How to get xix_i ?

e.g.

q1v=x1q1q1+0++0=x1q_1^\top v=x_1q_1q_1^\top+0+\cdots+0=x_1

i.e. (writing in the form of matrix)

[q1qn][x1xn]=vQx=v\begin{aligned} \begin{bmatrix} \vdots & & \vdots\\ q_1 & \cdots & q_n\\ \vdots & & \vdots \end{bmatrix} \begin{bmatrix} x_1 \\ \vdots \\ x_n \end{bmatrix} & = v \\ Qx & = v \end{aligned}

x=Q1v=Qvx=Q^{-1}v=Q^\top v

Fourier Series

f(x)=a0+a1cosx+b1sinx+a2cos2x+b2sin2x+f(x)=a_{0}+a_{1} \cos x+b_{1} \sin x+a_{2} \cos 2 x+b_{2} \sin 2 x+\cdots

$q_i areorthogonal,andare orthogonal, and \sin ,, \cos$ are orthogonal funtions. How come?

orthogonal funtions

like inner product, f=cosxf=\cos x and g=sinxg=\sin x are orthogonal funtions means:

fg=02πf(x)g(x)dx=12(sinx)202π=0f^{\top} g=\int_{0}^{2 \pi} f(x) g(x) d x=\left.\frac{1}{2}(\sin x)^{2}\right|_{0}^{2 \pi}=0

find the coefficient:

a1=fcos(x)coscos(x)=02πf(x)cos(x)dx02πcos(x)2dx=1π02πf(x)cos(x)dxa_1=\frac{f^\top \cos(x)}{\cos^\top \cos(x)}=\frac{\int_{0}^{2\pi}f(x)\cos(x) \mathrm{d}x}{\int_{0}^{2\pi}\cos(x)^2\mathrm{d}x}=\frac{1}{\pi}\int_{0}^{2\pi}f(x)\cos(x) \mathrm{d}x


MIT18.06 Linear Algebra (24)
https://yzzzf.xyz/2022/04/29/MIT16.08-lec24/
Author
Zifan Ying
Posted on
April 29, 2022
Licensed under