当前位置:网站首页>Diagonalization, power of a
Diagonalization, power of a
2022-07-25 18:18:00 【I don't know anything about Phuket】
One 、 Diagonalization of matrix and its conditions
1.2 Diagonalization of matrix
The square array A A A The eigenvector of is put into a new matrix S S S, So there are :
A S = A [ c 1 c 2 ⋯ c n ] = [ λ 1 c 1 ⋯ λ n c n ] = [ c 1 c 2 ⋯ c n ] [ λ 1 0 ⋯ 0 0 λ 2 ⋯ 0 ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ λ n ] = A Λ (1) \begin{aligned} AS&=A\begin{bmatrix}c_1&c_2&\cdots&c_n\end{bmatrix}\\&=\begin{bmatrix}\lambda_1c_1&\cdots&\lambda_nc_n\end{bmatrix}\\ &=\begin{bmatrix}c_1&c_2&\cdots& c_n\end{bmatrix}\begin{bmatrix}\lambda_1&0&\cdots&0\\0&\lambda_2&\cdots&0\\ \cdots&\cdots&\cdots&\cdots\\ \cdots&\cdots&\cdots&\lambda_n\end{bmatrix}=A\Lambda \end{aligned}\tag{1} AS=A[c1c2⋯cn]=[λ1c1⋯λncn]=[c1c2⋯cn]⎣⎡λ10⋯⋯0λ2⋯⋯⋯⋯⋯⋯00⋯λn⎦⎤=AΛ(1)
There is a :
A S = S Λ (2) AS=S\Lambda\tag{2} AS=SΛ(2)
Because eigenvectors are linearly independent , therefore :
S − 1 A S = Λ (3) S^{-1}AS=\Lambda\tag{3} S−1AS=Λ(3)
The following is the equivalent expression , That is the important formula of matrix diagonalization in our lesson :
A = S Λ S − 1 (4) A=S\Lambda S^{-1}\tag{4} A=SΛS−1(4)
This is another method of matrix decomposition : A matrix is divided into a matrix composed of eigenvalue vectors and a matrix composed of eigenvalues . Such decomposition , It makes it feasible for us to solve the power of the matrix .
from (3) There are the following conclusions :
A x = λ x A 2 x = λ A x = λ 2 x ⋯ A n x = λ n x (5) Ax=\lambda x\\ A^2x=\lambda Ax=\lambda^2x\\ \cdots\\ A^{n}x=\lambda^nx\tag{5} Ax=λxA2x=λAx=λ2x⋯Anx=λnx(5)
from (4) Yes :
A 2 = S Λ S − 1 S Λ S − 1 = S Λ 2 S − 1 A k = S Λ k S − 1 (6) A^2=S\Lambda S^{-1}S\Lambda S^{-1}=S\Lambda^2S^{-1}\\ A^k=S\Lambda^kS^{-1}\tag{6} A2=SΛS−1SΛS−1=SΛ2S−1Ak=SΛkS−1(6)
The existence of eigenvalues gives us an expression for calculating the power of a matrix , Be careful Matrix diagonalization condition yes : There is n n n Two independent eigenvectors .
Theorem : hypothesis A A A There is n n n Two independent eigenvectors , When k → ∞ k\rightarrow\infin k→∞, A k → 0 A^k\rightarrow0 Ak→0 Is the condition of :
∀ ∣ λ i ∣ < 1 (7) \forall \quad \vert\lambda_i\vert<1\tag{7} ∀∣λi∣<1(7)
Such a matrix A A A It is stable. .
Since there are so many good properties of diagonalization , So what kind of matrix can be diagonalized ? Here is the conclusion :
A A A There must be n n n Two independent vectors ( Or it can be diagonalized (diagnalizable)), If the following conditions are met :
- all λ \lambda λ Are different ( That is, there is no repetition λ \lambda λ, The textbook has proof )
Whether repeated eigenvalues must not be diagonalized ? answer : Look at the situation . The following is the two clock situation .
Example 1: Unit matrix I = [ 1 0 0 0 1 0 0 0 1 ] I=\begin{bmatrix}1&0&0\\0&1&0\\0&0&1\end{bmatrix} I=⎣⎡100010001⎦⎤, There are three eigenvalues , All are 1, But all vectors are their eigenvectors , Take whatever you like 3 Just one .
Example 2: For a triangular matrix A = [ 2 1 0 2 ] A=\begin{bmatrix}2&1\\0&2\end{bmatrix} A=[2012] The eigenvalue of is 2 and 2, Then find its eigenvector A − 2 I = [ 0 1 0 0 ] A-2I=\begin{bmatrix}0&1\\0&0\end{bmatrix} A−2I=[0010], The zero space of this matrix is a zero vector , The eigenvector is x 1 = [ 1 0 ] x_1=\begin{bmatrix}1\\0\end{bmatrix} x1=[10], Because I can't find 2 Linear independent vectors , No diagonalization .
Two 、 Application of diagonalization
If x i x_i xi yes A A A An eigenvector , that k x i kx_i kxi( k ≠ 0 k\neq 0 k=0) It's also its eigenvector .
A vector u 0 u_0 u0 It's a matrix A A A Eigenvector x i x_i xi The linear combination of :
u 0 = c 1 x 1 + c 2 x 2 + ⋯ + c n x n (8) u_0=c_1x_1+c_2x_2+\dots+c_nx_n\tag{8} u0=c1x1+c2x2+⋯+cnxn(8)
Matrix form :
u 0 = S c (9) u_0=Sc\tag{9} u0=Sc(9)
For everyone after u k u_k uk Through the general term :
u k + 1 = A u k (10) u_{k+1}=Au_k\tag{10} uk+1=Auk(10)
Find out , Concrete :
u 1 = A u 0 u 2 = A 2 u 0 ⋯ u k = A k u 0 (10) u_1=Au_0\\ u_2=A^2u_0\\ \cdots\\ u_k=A^ku_0\tag{10} u1=Au0u2=A2u0⋯uk=Aku0(10)
u 1 = A u 0 = c 1 λ 1 x 1 + c 2 λ 2 x 2 + ⋯ + c n λ n x n u 100 = A 100 u 0 = c 1 λ 1 100 x 1 + c 2 λ 2 100 x 2 + ⋯ + c n λ n 100 x n (11) u_1=Au_0=c_1\lambda_1x_1+c_2\lambda_2x_2+\cdots+c_n\lambda_nx_n\\ u_{100}=A^{100}u_0=c_1\lambda_1^{100}x_1+c_2\lambda_2^{100}x_2+\cdots+c_n\lambda_n^{100}x_n\tag{11} u1=Au0=c1λ1x1+c2λ2x2+⋯+cnλnxnu100=A100u0=c1λ1100x1+c2λ2100x2+⋯+cnλn100xn(11)
In matrix form :
u 0 = S c u 100 = u 0 A 100 = Λ 100 S c (12) u_0=Sc\\ u_{100}=u_0A^{100}=\Lambda^{100}Sc\tag{12} u0=Scu100=u0A100=Λ100Sc(12)
Here will be the initial vector u 0 u_0 u0 Expand into a linear combination of several eigenvectors , So we can deduce u 100 u_{100} u100 Result .
Fibonacci sequence (Fibonacci) Example :
0 , 1 , 1 , 2 , 3 , 5 , 8 , 13 ⋯ F 100 0,1,1,2,3,5,8,13\cdots F_{100} 0,1,1,2,3,5,8,13⋯F100
How to put the first 100 How to calculate Fibonacci number and its growth rate ?
answer : Growth rate 、 The convergence 、 Stability is determined by eigenvalues , The recurrence formula is given below :
F k + 2 = F k + 1 + F k (13) F_{k+2}=F_{k+1}+F_k\tag{13} Fk+2=Fk+1+Fk(13)
This equation is not only related to the previous value F k F_k Fk of , And also with the value of the previous F k F_k Fk of , It is a second-order difference equation .(13) You can also add a condition :
F k + 2 = F k + 1 + F k F k + 1 = F k + 1 (14) F_{k+2}=F_{k+1}+F_k\\ F_{k+1}=F_{k+1} \tag{14} Fk+2=Fk+1+FkFk+1=Fk+1(14)
In matrix form :
[ F k + 2 F k + 1 ] = [ 1 1 1 0 ] [ F k + 1 F k ] \begin{bmatrix}F_{k+2}\\F_{k+1}\end{bmatrix}=\begin{bmatrix}1&1\\1&0\end{bmatrix}\begin{bmatrix}F_{k+1}\\F_k\end{bmatrix} [Fk+2Fk+1]=[1110][Fk+1Fk]
Here we use a little trick to transform the second-order equation “ drop ” Is a first-order equation :
u k = [ F k + 1 F k ] u k + 1 = [ u k + 2 u k + 1 ] (15) u_k=\begin{bmatrix}F_{k+1}\\F_k\end{bmatrix}\\ u_{k+1}=\begin{bmatrix}u_{k+2}\\u_{k+1}\end{bmatrix}\tag{15} uk=[Fk+1Fk]uk+1=[uk+2uk+1](15)
Plug in (14) Yes :
u k + 1 = [ 1 1 1 0 ] u k (16) u_{k+1}=\begin{bmatrix}1&1\\1&0\end{bmatrix}u_k\tag{16} uk+1=[1110]uk(16)
from : ∣ A − λ I ∣ = − λ 2 − λ + 1 = 0 \vert A-\lambda I\vert=-\lambda^2-\lambda+1=0 ∣A−λI∣=−λ2−λ+1=0 The eigenvalues can be obtained as :
λ 1 = 1 + 5 2 ≈ 1.618 λ 2 = 1 − 5 2 ≈ 0.618 (17) \lambda_1=\frac{1+\sqrt{5}}{2}\approx1.618\\ \lambda_2=\frac{1-\sqrt{5}}{2}\approx0.618\tag{17} λ1=21+5≈1.618λ2=21−5≈0.618(17)
The corresponding eigenvector is :
[ 1 − λ 1 1 − λ ] x = 0 (18) \begin{bmatrix}1-\lambda&1\\1&-\lambda\end{bmatrix}x=0\tag{18} [1−λ11−λ]x=0(18)
x = [ λ 1 ] x=\begin{bmatrix}\lambda\\1\end{bmatrix} x=[λ1] yes (18) Eigenvector of , This is because :
A x = [ − λ 2 − λ + 1 0 ] = [ 0 0 ] (19) Ax=\begin{bmatrix}-\lambda^2-\lambda+1\\0\end{bmatrix}=\begin{bmatrix}0\\0\end{bmatrix}\tag{19} Ax=[−λ2−λ+10]=[00](19)
This is because − λ 2 − λ + 1 = 0 -\lambda^2-\lambda+1=0 −λ2−λ+1=0 Exactly equal to the characteristic root equation .
x 1 = [ 1.618 1 ] x 2 = [ 0.618 1 ] (20) x_1=\begin{bmatrix}1.618\\1\end{bmatrix}\quad x_2=\begin{bmatrix}0.618\\1\end{bmatrix}\tag{20} x1=[1.6181]x2=[0.6181](20)
In order to take advantage of the previous conclusion , We need to set the initial value u 0 u_0 u0 use A A A Linear combination representation of eigenvectors of matrix , Suppose this linear combination is c 0 c_0 c0 and c 1 c_1 c1
u 0 = c 1 x 1 + c 0 x 2 (21) u_0=c_1x_1+c_0x_2\tag{21} u0=c1x1+c0x2(21)
It doesn't matter what the result is . Let's review the whole problem-solving logic :
- Convert the second-order difference equation into the first-order difference equation , The purpose is to apply the power characteristics of the first-order difference equation
- Determine that the obtained eigenvalues and eigenvectors can be used to represent u 0 u_0 u0
- If the second step exists , So there are u k = A 100 u 0 = c 1 λ 1 100 x 1 + c 2 λ 2 100 x 2 + ⋯ + c n λ n 100 x n u_k=A^{100}u_0=c_1\lambda_1^{100}x_1+c_2\lambda_2^{100}x_2+\cdots+c_n\lambda_n^{100}x_n uk=A100u0=c1λ1100x1+c2λ2100x2+⋯+cnλn100xn
Because this is only a second-order matrix , therefore :
u k = c 1 λ 1 100 x 1 + c 2 λ 2 100 x 2 (22) u_k=c_1\lambda_1^{100}x_1+c_2\lambda_2^{100}x_2\tag{22} uk=c1λ1100x1+c2λ2100x2(22)
For this example :
u 100 = c 1 ( 0.618 ) 100 + c 2 ( 1.618 ) 100 (23) u_{100}=c_1(0.618)^{100}+c_2(1.618)^{100}\tag{23} u100=c1(0.618)100+c2(1.618)100(23)
Simple analysis , As the order increases , The eigenvalue is less than 1 Of (0.618) The impact will be smaller and smaller , So the eigenvalue of the decisive factor is the larger one (1.618).
边栏推荐
- Keil5 "loading PDSC debug description failed for STMicroelectronics stm32hxxxxxxx" solution
- NC78 反转链表
- Unittest framework application
- Basic operation of bidirectional linked list
- Li Kai: the interesting and cutting-edge audio and video industry has always attracted me
- Number two 2010 real test site
- PHP memory management mechanism and garbage collection mechanism
- Related operations of figure
- Oracle uses impdp import to report an error: ora-39001: invalid parameter value ora-39000: dump file description error ora-39088: file name cannot contain path description
- Is there any inconspicuous but profitable sideline?
猜你喜欢

Cloud XR面临的问题以及Cloud XR主要应用场景

这是一张机器&深度学习代码速查表

What scenarios have rust, which is becoming more and more mature, applied?

How to choose digital twin visualization platform

网易严选库存中心设计实践

ORB_SLAM3复现——上篇

Stm8s003f3 internal flash debugging

Introduction to cloud XR and development opportunities of cloud XR in 5g Era

二叉树的相关操作

c语言---25 扫雷游戏
随机推荐
testng执行顺序的3中控制方法
Use of stm8s003f3 UART
图的相关操作
What scenarios have rust, which is becoming more and more mature, applied?
工程师必看的示波器探头安全使用说明书
MySQL lost the previous 0 after the decimal number type select
Problems faced by cloud XR and main application scenarios of cloud XR
The use of invocationcount, threadpoolsize, timeout of concurrent tests in TestNG
Good news! Ruiyun technology was awarded the member unit of 5g integrated application special committee of "sailing on the sea"
How to choose digital twin visualization platform
大话DevOps监控,团队如何选择监控工具?
Cloud VR: the next step of virtual reality specialization
遍历数组的方法有哪些?for循环 forEach for/in for/of map的性能又有什么差别 该如何选择?
NC15 求二叉树的层序遍历
CH582 BLE 5.0 使用 LE Coded 广播和连接
"Deprecated gradle features were used in this build, making it incompatible with gradle 6.0" problem solving
国际权威认可!OceanBase入选Forrester Translytical数据平台报告
Tme2022 campus recruitment background development / operation development / business operation and maintenance / application development written examination (I) a little self analysis of programming q
PHP memory management mechanism and garbage collection mechanism
LeetCode 101. 对称二叉树 && 100. 相同的树 && 572. 另一棵树的子树