Linear Algebra Lecture 2

最新推荐文章于 2024-10-13 15:55:00 发布

Mummu

最新推荐文章于 2024-10-13 15:55:00 发布

阅读量192

点赞数

分类专栏： linear algebra 文章标签： linear algebra 线性代数算法基础

本文链接：https://blog.csdn.net/mummu/article/details/79054303

版权

linear algebra 专栏收录该内容

10 篇文章 2 订阅

订阅专栏

Linear Algebra Lecture 2

1. Elimination

2. Back-substitution

3. Elimination matrices

4. Multiply matrices

Example 1 :

To solve :

⎧ ⎩ ⎨ x + 2 y + z = 2 3 x + 8 y + z = 12 4 y + z = 2

$\begin{cases} x+2y+z =2\\ 3x+8y+z=12 \\ 4y+z=2 \end{cases}$

Here is a system of equations, three equations and three unknowns. Ax = b
Change the equations into matrix form:
$A = \begin{bmatrix} 1 & 2 & 1 \\ 3 & 8 & 1 \\0 & 4 & 1 \end{bmatrix}$

Elimination steps :
1.Accept the first equation is okay, multiply that equation by the right number, and going to subtract it from the second equation. With the purpose to knock out the x part of equation two, to eliminate x.

2.Find the first pivot(主元), won’t change the first row, because it’s the pivot row. Then find the multiplier, 3 times first equation will knock out that 3 in the second equation.

$\begin{bmatrix} 1 & 2 & 1 \\ 3 & 8 & 1 \\0 & 4 & 1 \end{bmatrix}\to\begin{bmatrix} 1 & 2 & 1 \\ 0 & 2 & -2 \\0 & 4 & 1 \end{bmatrix}$

3.We have eliminated x, we are down to two equations in y and z, find the second pivot 2, then eliminate y.

$\begin{bmatrix} 1 & 2 & 1 \\ 0 & 2 & -2 \\0 & 4 & 1 \end{bmatrix} \to \begin{bmatrix} 1 & 2 & 1 \\ 0 & 2 & -2 \\0 & 0 & 5 \end{bmatrix}$

4.Then the matrix is an upper triangular matrix(上三角矩阵), call U. The whole purpose of elimination was to get from A to U.

How could the elimination have failed?

If the first number was 0, there were no x in first equation. Then need switch rows, if there is a zero in the pivot position, need to exchange rows.

How we identify failure?
There’s temporary failure when we can do a row exchange and get out of it or there’s complete failure when we get a 0 and below.
There wouldn’t have been all pivots. The matrix would have not been invertible(矩阵不可逆).

Back substitution

Now bring the right-hand side as an extra column.

$\begin{bmatrix} 1 & 2 & 1 & 2\\ 3 & 8 & 1 & 12\\0 & 4 & 1 &2\end{bmatrix}$

This is called augmented matrix(增广矩阵). And we should do the same step to the right-hand side column.

$\begin{bmatrix} 1 & 2 & 1 & 2\\ 3 & 8 & 1 & 12\\0 & 4 & 1 &2\end{bmatrix} \to\begin{bmatrix} 1 & 2 & 1 &2 \\ 0 & 2 & -2 & 6 \\0 & 4 & 1 & 2\end{bmatrix} \to \begin{bmatrix} 1 & 2 & 1 &2\\ 0 & 2 & -2 & 6\\0 & 0 & 5 & -10\end{bmatrix}$

Then we need to do back substitution.

⎧ ⎩ ⎨ x + 2 y + z = 2 2 y - 2 z = 6 5 z = - 10

$\begin{cases} x+2y+z =2\\ 2y-2z=6 \\ 5z=-10 \end{cases}$

Then easy to solve the equations. Here is the back substitution(回代).

⎧ ⎩ ⎨ x = 2 y = 1 z = - 2

$\begin{cases} x =2\\ y=1 \\ z = -2 \end{cases}$
Back substitution is the simple step solving the equations in reverse order because the system is triangular now.

Elimination matrices

Row operations
Matrix times a columns is a column.
A row times a matrix is a row.
When as we do matrix multiplication, we got a linear combination of rows.

Elimination matrices(消元矩阵)
What’s the matrix that does the first step, subtracts 3 of equation one from equation two and leave the other rows the same?
$\begin{bmatrix}\vdots&\vdots&\vdots\end{bmatrix}\begin{bmatrix} 1 & 2 & 1 \\ 3 & 8 & 1 \\0 & 4 & 1 \end{bmatrix} = \begin{bmatrix} 1 & 2 & 1 \\ 0 & 2 & -2 \\0 & 4 & 1 \end{bmatrix}$

We are just doing this step changes row two, so actually, row one is not changing.
The first row of the matrix will be $\begin{bmatrix} 1 & 0 &0 \end{bmatrix}$ , because that’s just the right thing that takes one of the first row and none of the other rows. And the last row is $\begin{bmatrix} 0 &0 & 1\end{bmatrix}$ , takes one of the third row and none of the other rows.
The second row is a matrix $\begin{bmatrix} -3 &1 & 0\end{bmatrix}$ , means that takes minus 3 of row one plus the row two.
$\begin{bmatrix}1 &0 & 0\\ -3 &1 & 0\\0&0&1\end{bmatrix}\begin{bmatrix} 1 & 2 & 1 \\ 3 & 8 & 1 \\0 & 4 & 1 \end{bmatrix} = \begin{bmatrix} 1 & 2 & 1 \\ 0 & 2 & -2 \\0 & 4 & 1 \end{bmatrix}$

What matrix does nothing?
$\begin{bmatrix} 1& 0 &0 \\0 &1 &0\\0 &0 & 1\end{bmatrix}$
The identity matrix(单位矩阵), it just multiplies everything and leaves it where it is, like the number 1.

Elementary matrix(初等矩阵)
$E_{21}$ is the matrix we needed to fix the $(2,1)$ position to be 0.

Step 2 is subtract 2 times row 2 from row 3
$E_{32} \to \begin{bmatrix}1 &0 & 0\\ 0 &1 & 0\\0&-2&1\end{bmatrix}\begin{bmatrix} 1 & 2 & 1 \\ 0 & 2 & -2 \\0 & 4 & 1 \end{bmatrix} = \begin{bmatrix} 1 & 2 & 1 \\ 0 & 2 & -2 \\0 & 0 & 5 \end{bmatrix}$

The elimination matrices that take each step.Put those together into a matrix that does it all.

$E_{32}(E_{21}A)=U \to (E_{32}E_{21})A=U$

What matrix takes me from A to U? $E$

Permutation

There’s another type of elementary matrix, the matrix that exchanges two rows, it’s called a permutation matrix(置换矩阵).
Permutation = Exchange rows 1 and 2
$P\to\begin{bmatrix}0&1\\ 1&0\end{bmatrix}\begin{bmatrix} a&b \\ c&d\end{bmatrix} = \begin{bmatrix} c&d\\a&b \end{bmatrix}$

If need to do column exchanges,
$\begin{bmatrix} a&b \\ c&d\end{bmatrix} \begin{bmatrix}0&1\\ 1&0\end{bmatrix}= \begin{bmatrix} b&a\\d&c \end{bmatrix}$

In short, to do column operations, the permutation matrix multiplies on the right, to do row operations , it multiplies on the left.

Inverse Matrix

Inverses

$\begin{bmatrix} 1&0&0\\3&1&0\\0&0&1 \end{bmatrix}\begin{bmatrix} 1&0&0\\-3&1&0\\0&0&1 \end{bmatrix} = \begin{bmatrix} 1&0&0\\0&1&0\\0&0&1 \end{bmatrix}$