「統計解析のための線形代数」復習筆記12

矩陣乘法運算

矩陣乘法定義

Theorem 1 (matrix multiplication) 兩個矩陣 \(A, B\) ,只有 \(A\) 的列數和 \(B\) 的行數相等(這種特徵又被稱爲:矩陣 \(A,B\) 可整合的,conformable)時,才有定義:\(AB\)\(AB\) 則爲新的矩陣,類型爲 \(A\) 的行數, \(B\)的列數。即:\(A_{k\times l}, \; B_{m\times n}\)\(l=m\) 時才能計算乘積: \(AB_{k\times n}\)
  1. \(A_{2\times3}=\left( \begin{array}{c} 4 & 6 & 8\\ 2 & 1 & 3\\ \end{array} \right),\; B_{3\times2}=\left( \begin{array}{c} 0 & 8\\ 2 & -1\\ 9 & 4 \\ \end{array} \right)\) 時,
    \(A\)的列數” \(=\)\(B\) 的行數” \(= 3\),因此積 \(AB\) 被定義,類型是 \((2,2)\)
    \(B\)的列數” \(=\)\(A\) 的行數” \(= 2\),因此積 \(BA\) 被定義,類型是 \((3,3)\)

  2. \(C_{3\times2}=\left( \begin{array}{c} 4 & 2 \\ 5 & 6 \\ 7 & 3 \end{array} \right),\; D_{2\times4}=\left( \begin{array}{c} 2 & 0 & 9 & -1 \\ 4 & 7 & 6 & 5 \\ \end{array} \right)\) 時,
    \(C\)的列數” \(=\)\(D\) 的行數” \(= 2\),因此積 \(CD\) 被定義,類型是 \((3,4)\)
    \(D\)的列數”\(= 4\),“\(C\) 的行數” \(= 3\),因此積 \(DC\) 不被定義,\(DC\) 不可整合。

矩陣的積的向量表達形式

矩陣也可以被看做是一個個相同類型(大小,方向)的向量組成。那麼當,下面\(A, B\)兩個矩陣滿足:\(A\) 的行向量的維度,和\(B\)的列向量的維度相等時,\(AB\)被定義。

\(A=\left( \begin{array}{c} \underline{a}_1^t\\ \underline{a}_2^t\\ \vdots\\ \underline{a}_k^t \end{array} \right), \; B=(\underline{b}_1,\underline{b}_2,\cdots,\underline{b}_n)\)

\(AB=A(\underline{b}_1,\underline{b}_2,\cdots,\underline{b}_n)=(A\underline{b}_1,A\underline{b}_2,\cdots,A\underline{b}_n)\)

或者:

\(AB=\left( \begin{array}{c} \underline{a}_1^t\\ \underline{a}_2^t\\ \vdots\\ \underline{a}_k^t \end{array} \right)B=\left( \begin{array}{c} \underline{a}_1^tB\\ \underline{a}_2^tB\\ \vdots\\ \underline{a}_k^tB \end{array} \right)\)

特殊情況當\(A\)僅有一個行向量 \(\underline{a}^t\) 時,
\(AB=\underline{a}^tB=\underline{a}^t(\underline{b}_1,\underline{b}_2,\cdots,\underline{b}_n)=(\underline{a}^t\underline{b}_1,\underline{a}^t\underline{b}_2,\cdots,\underline{a}^t\underline{b}_n)\)

矩陣的積的成分

Theorem 2 (matrix multiplication component) 兩個矩陣 \(A_{k\times l}, \; B_{m\times n}\) 的積有被定義時,矩陣 \(AB_{k\times n}\) 的任意成分\((i,j)\),被定義爲:\(A\) 的第 \(i\) 行向量與 \(B\)的第 \(j\) 列向量的內積”

故:
\(AB\)\((1,1)\) 成分是,“\(A\) 的第 \(1\) 行向量與 \(B\)的第 \(1\) 列向量的內積”
\(AB\)\((1,2)\) 成分是,“\(A\) 的第 \(1\) 行向量與 \(B\)的第 \(2\) 列向量的內積”
\(\vdots\)
\(AB\)\((k,n)\) 成分是,“\(A\) 的第 \(k\) 行向量與 \(B\)的第 \(ns\) 列向量的內積”

練習

  1. \(A=\left( \begin{array}{c} 1 & 2 \\ 3 & 4 \\ \end{array} \right), B=\left( \begin{array}{c} 4 & 1 \\ 5 & 2 \\ \end{array} \right)\)
    \(AB=\left( \begin{array}{c} 1\times4+2\times5 & 1\times1+2\times2 \\ 3\times4+4\times5 & 3\times1+4\times2 \\ \end{array} \right)\\ \;\;\;\;\;=\left( \begin{array}{c} 14 & 5 \\ 32 & 11 \\ \end{array} \right)\)
    \(BA=\left( \begin{array}{c} 4\times1+1\times3 & 4\times2+1\times4 \\ 5\times1+2\times3 & 5\times2+2\times4 \\ \end{array} \right)\\ \;\;\;\;\;=\left( \begin{array}{c} 7 & 12 \\ 11 & 18 \\ \end{array} \right)\)
    注意: \(AB\neq BA\)
    另外:\(AA=\left( \begin{array}{c} 1\times1+2\times3 & 1\times2+2\times4 \\ 3\times1+4\times3 & 3\times2+4\times4 \\ \end{array} \right)\\ \;\;\;\;=\left( \begin{array}{c} 7 & 10 \\ 15 & 22 \\ \end{array} \right)=AA^2\)

  2. \(A=\left( \begin{array}{c} 1 & 2 \\ -2 & -4 \\ \end{array} \right), B=\left( \begin{array}{c} 6 & -2 \\ -3 & 1 \\ \end{array} \right)\)
    \(AB=\left( \begin{array}{c} 1\times6+2\times(-3) & 1\times(-2)+2\times1 \\ (-2)\times6+(-4)\times(-3) & (-2)\times(-2)+(-4)\times1 \\ \end{array} \right)\\ \;\;\;\;=\left( \begin{array}{c} 0 & 0 \\ 0 & 0 \\ \end{array} \right)=\Large 0\)
    這裏出現了非零矩陣相乘爲零矩陣 \(\large 0\)的例子。

  3. \(A_{2\times3}=\left( \begin{array}{c} a_{11} & a_{12} & a_{13}\\ a_{21} & a_{22} & a_{23}\\ \end{array} \right)=\left( \begin{array}{c} \underline{a}_1^t\\ \underline{a}_2^t\\ \end{array} \right)=(\underline{c}_1,\underline{c}_2,\underline{c}_3); \\ B_{3\times2}=\left( \begin{array}{c} b_{11} & b_{12}\\ b_{21} & b_{22}\\ b_{31} & b_{32}\\ \end{array} \right)=(\underline{b}_1,\underline{b}_2)=\left( \begin{array}{c} \underline{d}^t_1\\ \underline{d}^t_2\\ \underline{d}^t_3\\ \end{array} \right)\)
    \(AB_{2\times2}\)\(=\left( \begin{array}{c} \sum\limits_{k=1}^3a_{1k}b_{k1} &\sum\limits_{k=1}^3a_{1k}b_{k2} \\ \sum\limits_{k=1}^3a_{2k}b_{k1} &\sum\limits_{k=1}^3a_{2k}b_{k2} \\ \end{array} \right)\\=\left( \begin{array}{c} \underline{a}^t_1\underline{b}_1 & \underline{a}^t_1\underline{b}_2 \\ \underline{a}^t_2\underline{b}_1 & \underline{a}^t_2\underline{b}_2 \\ \end{array} \right)\)
    \(BA_{3\times3}\)\(=\left( \begin{array}{c} \sum\limits_{k=1}^2b_{1k}a_{k1} & \sum\limits_{k=1}^2b_{1k}a_{k2} & \sum\limits_{k=1}^2b_{1k}a_{k3} \\ \sum\limits_{k=1}^2b_{2k}a_{k1} & \sum\limits_{k=1}^2b_{2k}a_{k2} & \sum\limits_{k=1}^2b_{2k}a_{k3} \\ \sum\limits_{k=1}^2b_{3k}a_{k1} & \sum\limits_{k=1}^2b_{3k}a_{k2} & \sum\limits_{k=1}^2b_{3k}a_{k3} \end{array} \right)\\ =\left( \begin{array}{c} \underline{d}^t_1\underline{c}_1 & \underline{d}^t_1\underline{c}_2 & \underline{d}^t_1\underline{c}_3 \\ \underline{d}^t_2\underline{c}_1 & \underline{d}^t_2\underline{c}_2 & \underline{d}^t_2\underline{c}_3 \\ \underline{d}^t_3\underline{c}_1 & \underline{d}^t_3\underline{c}_2 & \underline{d}^t_3\underline{c}_3 \\ \end{array} \right)\)

  4. \(\underline{x}=\left( \begin{array}{c} x_1\\ x_2\\ \vdots\\ x_n\\ \end{array} \right)\) 時,
    \(\underline{x}\underline{x}^t=\left( \begin{array}{c} x_1^2 & x_1x_2 & \cdots & x_1x_n \\ x_2x_1 & x_2^2 & \cdots & x_2x_n \\ \vdots & \vdots & \ddots & \vdots \\ x_nx_1 & x_nx_2 & \cdots & x_n^2 \\ \end{array} \right)\)
    \(\underline{x}^t\underline{x}=\sum\limits_{i=1}^nx_i^2\)

  5. \(\underline{a}=\left( \begin{array}{c} a_1\\ a_2\\ a_3\\ \end{array} \right), \underline{b}=\left( \begin{array}{c} b_1\\ b_2\\ b_3\\ \end{array} \right), \underline{c}=\left( \begin{array}{c} c_1\\ c_2\\ c_3\\ \end{array} \right)\\ A=\left( \begin{array}{c} a_1 & b_1 & c_1 \\ a_2 & b_2 & c_2 \\ a_3 & b_3 & c_3 \\ \end{array} \right)\)
    \(\underline{a}\underline{a}^t+\underline{b}\underline{b}^t+\underline{c}\underline{c}^t\\ \;\;=\left( \begin{array}{c} a_1^2 & a_1a_2 & a_1a_3 \\ a_2a_1 & a_2^2 & a_2a_3 \\ a_3a_1 & a_3a_2 & a_3^2 \\ \end{array} \right)\\ \;\;\;\;\;\;+\left( \begin{array}{c} b_1^2 & b_1b_2 & b_1b_3 \\ b_2b_1 & b_2^2 & b_2b_3 \\ b_3b_1 & b_3b_2 & b_3^2 \\ \end{array} \right)\\ \;\;\;\;\;\;+\left( \begin{array}{c} c_1^2 & c_1c_2 & c_1c_3 \\ c_2c_1 & c_2^2 & c_2c_3 \\ c_3c_1 & c_3c_2 & c_3^2 \\ \end{array} \right)\\ =\left( \begin{array}{c} a_1 & b_1 & c_1 \\ a_2 & b_2 & c_2 \\ a_3 & b_3 & c_3 \\ \end{array} \right)\left( \begin{array}{c} a_1 & a_2 & a_3 \\ b_1 & b_2 & b_3 \\ c_1 & c_2 & c_3 \\ \end{array} \right)=AA^t\)

  6. \(\underline{x}=\left( \begin{array}{c} x_1\\ x_2\\ \vdots\\ x_n\\ \end{array} \right), \; \underline{\frac{1}{n}}=\left( \begin{array}{c} \frac{1}{n}\\ \frac{1}{n}\\ \cdots\\ \frac{1}{n}\\ \end{array} \right)\) 時, \(\underline{x}^t\underline{\frac{1}{n}}=\sum\limits_{i=1}^nx_i\cdot\frac{1}{n}=\bar{x}\) (\(\underline{x}\)平均值)
    將這樣的 \(\bar{x}\) 寫成 \(n\) 個的橫向量:\((\bar{x},\bar{x},\cdots,\bar{x})\)
    這個向量如果寫成展開的形式就是:
    \((\bar{x},\bar{x},\cdots,\bar{x})=(\underline{x}^t\underline{\frac{1}{n}}, \underline{x}^t\underline{\frac{1}{n}}, \cdots,\underline{x}^t\underline{\frac{1}{n}})\\ \;\;\;\;\;\;=\underline{x}^t(\underline{\frac{1}{n}},\underline{\frac{1}{n}},\cdots,\underline{\frac{1}{n}})\\ \;\;\;\;\;\;=(x_1,x_2,\cdots,x_n)\left( \begin{array}{c} \frac{1}{n} & \frac{1}{n} & \cdots & \frac{1}{n} \\ \frac{1}{n} & \frac{1}{n} & \cdots & \frac{1}{n} \\ \vdots & \vdots & \ddots & \vdots \\ \frac{1}{n} & \frac{1}{n} & \cdots & \frac{1}{n} \\ \end{array} \right)\)

矩陣積的性質

  • 矩陣 \(AB\) 有定義時, \(BA\) 並不一定就有定義。無法整合時就沒有定義。
  • \(AB=BA\) 時,\(A, B\) 被稱爲可交換 commutative交換可能矩陣
  • \(AB, BA\) 都有定義時,也不一定就滿足 \(AB=BA\)。也就是說,多數情況下, \(AB\neq BA\)。爲了區分二者,\(AB\) 被稱爲 \(A\) 從右往左乘 \(B\) (postmultiplication of \(A\) by \(B\))\(BA\) 被稱爲 \(A\) 從左往右乘 \(B\) (postmultiplication of \(B\) by \(A\))
  • 相似的, \(AC=BC\) 時,應該理解爲: 等式\(A=B\)兩邊同時從右往左乘 \(C\)
    \(CA=CB\) 就是:等式\(A=B\)兩邊同時從左往右乘\(C\)
  • 即使 \(A\neq\Large 0\)\(B\neq\Large 0\)\(AB\) 也有可能等於 \(\Large 0\) (零矩陣),此時我們說, \(A, B\)零因子 (zero divisor)
Avatar
王 超辰 - Chaochen Wang
Real World Evidence Scientist

All models are wrong, but some are useful.

comments powered by Disqus

Related