其他連結

WikiWand

LU

ptt 對角線討論 \( L \): lower matrix
\( U \): upper matrix
\( A=LU \) \( \leadsto \) LU-decomposition
\( A\in F^{m\times n} \)
\( L\in F^{m\times m}, U\in F^{m\times n} \)

\( x\in F^{n\times 1}, b\in F^{m\times 1} \)
\( Ax=b \)
\( \Rightarrow LUx=b \)

LDU

D: Diagonal matrix
\( A\in F^{m\times n} \)
\( L\in F^{m\times m}, D\in F^{m\times m}, U\in F^{m\times n} \)

\( A=LU_1= \begin{bmatrix} 1 & 2 & 3 \\ 4 & -6 & -6 \\ -7 & -7 & 9 \end{bmatrix} = \begin{bmatrix} 1 & 0 & 0 \\ 4 & 1 & 0 \\ 7 & 1/2 & 1 \end{bmatrix} \begin{bmatrix} 1 & 2 & 3 \\ 0 & -14 & -18 \\ 0 & 0 & 21 \end{bmatrix} \)

\( =LDU = \begin{bmatrix} 1 & 0 & 0 \\ 4 & 1 & 0 \\ 7 & 1/2 & 1 \end{bmatrix} \begin{bmatrix} 1 & 0 & 0 \\ 0 & -14 & 0 \\ 0 & 0 & 21 \end{bmatrix} \begin{bmatrix} 1/1 & 2/1 & 3/1 \\ 0 & (-14)/(-14) & (-18)/(-14) \\ 0 & 0 & 21/21 \end{bmatrix} \)

\( = \begin{bmatrix} 1 & 0 & 0 \\ 4 & 1 & 0 \\ 7 & 1/2 & 1 \end{bmatrix} \begin{bmatrix} 1 & 0 & 0 \\ 0 & -14 & 0 \\ 0 & 0 & 21 \end{bmatrix} \begin{bmatrix} 1 & 2 & 3 \\ 0 & 1 & 9/7 \\ 0 & 0 & 1 \end{bmatrix} \)

Notice

\( E\in F^{m\times m} , A\in F^{m\times n} \)
\( E=R_kR_{k-1}\dots R_1 \)
\( \because R \text{ is inverse.} \)
\( \therefore E \text{ is inverse.} \)

\( A\xrightarrow[\quad\text{without permutation of row}\quad]{E} U \)

\( EA=U \)
\( L=E^{-1} \)
\( \Rightarrow (R_kR_{k-1}\dots R_1)A=U \)
\( A= (R_kR_{k-1}\dots R_1)^{-1}U = R_{1}^{-1}R_{2}^{-1}\dots R_{k}^{-1}U \)

Example

Find an LU-decomposition of \( A= \begin{bmatrix} 1 & 2 & 3 \\ 4 & -6 & -6 \\ -7 & -7 & 9 \end{bmatrix}. \)

solution

\( A\xrightarrow{r_{12}^{(-4)}, r_{13}^{(7)}, r_{23}^{(\frac{1}{2})}} \begin{bmatrix} 1 & 2 & 3 \\ 0 & -14 & -18 \\ 0 & 0 & 21 \end{bmatrix} =U \)

\( E= R_{23}^{(\frac{1}{2})}R_{13}^{(7)}R_{12}^{(-4)} \)
\( L=E^{-1} \)
\( =(R_{23}^{(\frac{1}{2})}R_{13}^{(7)}R_{12}^{(-4)})^{-1} \)
\( = (R_{12}^{(-4)})^{-1} (R_{13}^{(7)})^{-1} (R_{23}^{(\frac{1}{2})})^{-1} \)
\( = R_{12}^{(4)} R_{13}^{(-7)} R_{23}^{(-\frac{1}{2})} \)

\( = \begin{bmatrix} 1 & 0 & 0 \\ 4 & 1 & 0 \\ 0 & 0 & 1 \end{bmatrix} \begin{bmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \\ -7 & 0 & 1 \end{bmatrix} \begin{bmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & -1/2 & 1 \end{bmatrix} = \begin{bmatrix} 1 & 0 & 0 \\ 4 & 1 & 0 \\ -7 & -1/2 & 1 \end{bmatrix} \)


\( \ast \ast \ast \begin{bmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & -1/2 & 1 \end{bmatrix} \begin{bmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \\ -7 & 0 & 1 \end{bmatrix} \begin{bmatrix} 1 & 0 & 0 \\ 4 & 1 & 0 \\ 0 & 0 & 1 \end{bmatrix} = \begin{bmatrix} 1 & 0 & 0 \\ 4 & 1 & 0 \\ -9 & 1/2 & 1 \end{bmatrix} \neq \begin{bmatrix} 1 & 0 & 0 \\ 4 & 1 & 0 \\ -7 & -1/2 & 1 \end{bmatrix} \)

Application

Using LU-decomposition to solute linear-system equation.
\( Ax=b \)
\( \rightarrow LUx=b \)
let \( Ux=y \)
thus, \( Ly=b \)
Step 1: forward substitution solute y
Step 2: back substitution solute x

題目

台大 99 資工

Reference

黃子嘉-線性代數