MIT18.06学习笔记 - Lecture 5: Transposes and Permutations

这个系列文章是我重温Gilbert老爷子的线性代数在线课程的学习笔记。
Course Name：MIT 18.06 Linear Algebra
Text Book: Introduction to Linear Algebra
章节内容: 2.7

---

课程提纲
1. Transpose and Symmetric matrices
2. Permutation Matrices
3. $PA=LU$ Factorization with Row Exchanges

课程重点

Transpose and Symmetric matrices

Transpose of $A$ is denoted by $A^T$. The columns of $A^T$ are the rows of $A$: $(A^T)_{ij} = A_{ji}$
Rules for transposes: $(AB)^T = B^TA^T$ and $(A^{-1})^T = (A^T)^{-1}$.
Symmetric matrices: $A^T=A$.
$R^TR$ and $RR^T$ is always symmetric, $R$ is any matrix:

Symmetric matrices in elimination $A=LDL^T$
If $A=A^T$ is factored into $LDU$ with no row exchanges, then $U$ is exactly $L^T$. The symmetric factorization of a symmetric matrix is $A=LDL^T$:

Permutation Matrices

A permutation matrix $P$ has the rows of the identity $I$ in any order. $P^T$ is also a permutation matrix and product $P_1P_2$ is again a permutation matrix. There are $n!$ permutation matrices of order n:

$p^{-1}$ is also a permutation matrix and always the same as $p^T$: $p^{-1} = p^T$.
In all case, a single row exchange is its own inverse (transpose). If we repeat the exchange we are back to $I$. But for $P_{32}P_{21}$, the inverses go in opposite order: $P_{21}P_{32}$.

$PA=LU$ Factorization with Row Exchanges

If $A$ is invertible, a permutation $P$ will put its rows in the right order in advance (so that no exchanges are needed for $PA$) to factor $PA=LU$:

The matrix $PA$ has its rows in good order, and it factors as usual into $LU$: