Rechnerische Hintergründe des Matrix-Begriffes und Herleitung des Matrizenproduk
Von: Wario
Datum: Mo. 28. August 2023 06:26:27
Thema: Lineare Algebra

Rechnerische Hintergründe des Matrix-Begriffes und Herleitung des Matrizenproduktes $\begin{pmatrix} a_{11} & a_{12} & \cdots & a_{1n} \\ a_{21} & a_{22} & \cdots & a_{2n} \\ \vdots & \vdots & \ddots & \vdots \\ a_{m1} & a_{m2} & \cdots & a_{mn} \\ \end{pmatrix} =A$

In der Oberstufe oder im 1. Semester wird die Matrix zumeist als rechtecksförmiges Zahlen- bzw. Größenschema eingeführt; was sozusagen einer Nominaldefinition entspricht. Einen operativen Sinn erhalten Matrizen durch das Matrizenprodukt, für welches das übliche Berechnungsverfahren für die Einträge der Produktmatrix (etwa $A\cdot B$), nach dem Prinzip $\text{Zeile}\times\text{Spalte},$ gelehrt wird. Das Matrixprodukt kann somit gewissermaßen als Realdefinition der Matrix selbst verstanden werden. Im vorliegenden Artikel soll es um folgendes gehen: · Eine Matrix stellt (lediglich) eine Kurzschreibweise dar, mit der eine weit elementarere Vektoroperation ausgedrückt werden kann, und zwar die Linearkombination, das ist eine spezielle Vektorsumme. · Eine Linearkombination kann (als Definition) als Matrixprodukt $\text{Matrix} \times \text{Vektor} =\text{neuer Vektor}$ umgeschrieben werden. · Das allgemeine Matrixprodukt $\text{Multiplikanden-Matrix} \times \text{Multiplikator-Matrix} =\text{neue Matrix}$ kann aus dem zuvorgenannten Produkt $\text{Matrix} \times \text{Vektor}$ hergeleitet werden.

Übersicht.

· Wiederholung benötigter Rechenregeln und Definitionen. · Beispiel zur Veranschaulichung einer Linearkombination als Vektorkette. · Nebenrechnung zur Herleitung des Matrix-Begriffes. · Der Begriff Matrix und das Matrixprodukt $\text{Matrix}\times\text{Vektor}.$ · Falk-Schema. · Weitere Definitionen. · Das Produkt $\text{Multiplikanden-Matrix} \times \text{Multiplikator-Matrix}.$ · Das allgemeine Matrixprodukt. · Beispiel zur Veranschaulichung einer Matrixmultiplikation. · Quellen.


 

Wiederholung benötigter Rechenregeln und Definitionen.

Wir stellen uns auf den Standpunkt, dass wir nicht wissen, was eine Matrix oder ein Matrizenprodukt ist. (i) Unter der S-Multiplikation eines $n$-dimensionalen Vektors $ \vec{v} =\begin{pmatrix} v_1 \\ v_2 \\ \vdots \\ v_n \end{pmatrix} \in \mathbb{R}^n $ versteht man die Streckung (bzw. Stauchung) um einen Faktor $\lambda \in \mathbb{R}$, den sogenannten Streckfaktor, das heißt $ \lambda\cdot \vec{v} =\lambda\cdot \begin{pmatrix} v_1 \\ v_2 \\ \vdots \\ v_n \end{pmatrix} =\begin{pmatrix} \lambda\cdot v_1 \\ \lambda\cdot v_2 \\ \vdots \\ \lambda\cdot v_n \end{pmatrix}.$ (ii) Für die S-Multiplikation gilt das Kommutativgesetz (Vertauschungsregel): $ \lambda\cdot \vec{v} =\vec{v}\cdot \lambda.$ (Die Regel dient hier als Schreibweise, die später noch nützlich sein wird.) (iii) Die Summe zweier Vektoren berechnet man durch komponentenweise Addition: $ \begin{pmatrix} v_1 \\ v_2 \\ \vdots \\ v_n \end{pmatrix} +\begin{pmatrix} w_1 \\ w_2 \\ \vdots \\ w_n \end{pmatrix} =\begin{pmatrix} v_1+w_1 \\ v_2+w_2 \\ \vdots \\ v_n+w_n \end{pmatrix}.$ (iv) Eine Summe gestreckter Vektoren $ \lambda_1\cdot \vec{v}_1 +\lambda_2\cdot \vec{v}_2 +\dots +\lambda_k\cdot \vec{v}_k $ wird Linearkombination genannt. Anschaulich geometrisch erzeugt eine Linearkombination eine sogenannte Vektorkette, die einen neuen Vektor von Startpunkt zu Endpunkt liefert; siehe Beispiel im nächsten Abschnitt.
 

Beispiel zur Veranschaulichung einer Linearkombination als Vektorkette.

\def\lI{1.5} \def\lII{0.5} \def\lIII{3} \def\lIV{-1} \scalebox{0.625}{ \begin{tikzpicture}[ gridded, scale=1, >=latex, every path/.style={thick, ->}, ] \coordinate[label=] (S) at (0,0); \coordinate[label=] (A) at (2,2); \coordinate[label=] (As) at ($\lI*(A)$); %A \draw[->] (S) -- (A) node[midway, sloped, above]{$\vec{a}_1$}; \draw[->, blue,] ([xshift=1cm]S) -- ([xshift=1cm]As) node[midway, sloped, above]{$\lI\, \vec{a}_1$}; %B \begin{scope}[shift={(4,0)}] \coordinate[label=] (S) at (0,0); \coordinate[label=] (B) at (4,0); \coordinate[label=] (Bs) at ($\lII*(B)$); \draw[->] (S) -- (B) node[midway, sloped, above]{$\vec{a}_2$}; \draw[->, blue,] ([shift={(0,1)}]S) -- ([shift={(0,1)}]Bs) node[midway, sloped, above]{$\lII\, \vec{a}_2$}; \end{scope} %C \begin{scope}[shift={(9,0)}] \coordinate[label=] (S) at (0,0); \coordinate[label=] (C) at (0,1); \coordinate[label=] (Cs) at ($\lIII*(C)$); \draw[->] (S) -- (C) node[midway, left]{$\vec{a}_3$}; \draw[->, blue,] ([shift={(1,0)}]S) -- ([shift={(1,0)}]Cs) node[midway, right]{$\lIII\, \vec{a}_3$}; \end{scope} %D \begin{scope}[shift={(12,0)}] \coordinate[label=] (S) at (0,0); \coordinate[label=] (D) at (3,1); \coordinate[label=] (Ds) at ($\lIV*(D)$); \draw[->] (S) -- (D) node[midway, sloped, above]{$\vec{a_4}$}; \draw[->, blue,] ([shift={(3,2)}]S) -- ([shift={(3,2)}]Ds) node[midway, sloped, above]{$- \vec{a}_4$}; \end{scope} \begin{scope}[shift={(5,-8)}] \coordinate[label=] (S) at (0,0); \coordinate[label=] (A) at (2,2); \coordinate[label=] (As) at ($\lI*(A)$); \coordinate[label=] (B) at (4,0); \coordinate[label=] (Bs) at ($\lII*(B)$); \coordinate[label=] (C) at (0,1); \coordinate[label=] (Cs) at ($\lIII*(C)$); \coordinate[label=] (D) at (3,1); \coordinate[label=] (Ds) at ($\lIV*(D)$); \draw[blue] (S) coordinate(X0) -- (As) coordinate (X1) node[midway, sloped, above]{$\lI\, \vec{a}_1$}; \draw[blue] (X1) -- +(Bs) coordinate (X2) node[midway, sloped, above]{$\lII\, \vec{a}_2$}; \draw[blue] (X2) -- +(Cs) coordinate (X3) node[midway, right]{$\lIII\,\vec{a}_3$}; \draw[blue] (X3) -- +(Ds) coordinate (X4) node[midway, sloped, above]{$-\vec{a}_4$}; \draw[very thick, red] (S) -- (X4) node[midway, right, anchor=east]{ $\lI\, \vec{a}_1 +\lII\, \vec{a}_2 +\lIII\, \vec{a}_3 -\vec{a}_4$ }; \foreach \k in {0,...,4} \draw[fill=black!1] (X\k) circle(2.0pt); \end{scope} \end{tikzpicture} }
 

Nebenrechnung zur Herleitung des Matrix-Begriffes.

Es seien $m$-dimensionale Vektoren $ % \newcommand\a[1]{% \begin{pmatrix} a_{1#1} \\ a_{2#1} \\ \vdots \\ a_{m#1} \end{pmatrix} } % \vec{a}_1,~ \vec{a}_2,~ \dots~ \vec{a}_n \,\in \mathbb{R}^m $ in einer Anzahl $n$ vorgelegt, das heißt $ \vec{a}_{\color{blue} 1} =\a{{\color{blue} 1}},~ \vec{a}_2 =\a{2},~ \dots,~ \vec{a}_n =\a{n}; $ sowie gleichviele reelle Zahlen $x_1,~ x_2, \dots,~ x_n \in\mathbb{R}$. Es sei $\vec{y}$ jene Linearkombination, die sich mittels oben genannter Größen bilden lässt, das heißt $ \vec{y} =x_1\, \vec{a}_1 +x_2\, \vec{a}_2 +\dots +x_n\, \vec{a}_n.$ Für die Koordinaten $ \begin{pmatrix}y_1 \\ y_2 \\ \vdots \\ y_m \end{pmatrix} =\vec{y}$ wird dann $\newcommand\dotline[1][]{% \ldots \ldots \ldots \ldots \ldots \ldots \ldots \ldots \ldots \ldots \ldots \ldots \ldots~#1} % \def\y{\begin{pmatrix} y_1 \\ y_2 \\ \vdots \\ y_m \end{pmatrix}} % \newcommand\ya[1]{% \begin{pmatrix} a_{1#1}x_#1 \\ a_{2#1}x_#1 \\ \vdots \\ a_{m#1}x_#1 \end{pmatrix} } % \def\yR{\begin{pmatrix} a_{11} x_1 +a_{12} x_2 +\dots +a_{1n} x_n \\ a_{21} x_1 +a_{22} x_2 +\dots +a_{2n} x_n \\ \dots \dots \dots \dots \dots \dots \dots \dots \dots \\ a_{m1} x_1 +a_{m2} x_2 +\dots +a_{mn} x_n \\ \end{pmatrix}} % \def\A{\begin{pmatrix} a_{11} & a_{12} & \cdots & a_{1n} \\ a_{21} & a_{22} & \cdots & a_{2n} \\ \vdots & \vdots & \ddots & \vdots \\ a_{m1} & a_{m2} & \cdots & a_{mn} \\ \end{pmatrix}} % % % \begin{array}{l l} \vec{y} &=x_1\, \vec{a}_1 +x_2\, \vec{a}_2 +\dots +x_n\, \vec{a}_n \\[1.5em] &=x_1\a{1} +x_2\a{2} +\dots +x_n\a{n} \\[1.5em] &\overset{\textsf{(ii)}}{=} \a{1}x_1 +\a{2}x_2 +\dots +\a{n}x_n \dotline[\textsf{(1)}] \\[1.5em] &=\ya{1} + \ya{2} +\dots +\ya{n} \\[1.5em] &\overset{\textsf{(iii)}}{=} \yR =\y =\vec{y} \dotline[\textsf{(2)}] \end{array}$
 

Der Begriff Matrix und das Matrixprodukt $\text{Matrix}\times\text{Vektor}.$

Zu dem Ausgangspunkt $\textsf{(1)}$ im vorangehenden Abschnitt soll nun eine Kurzschreibweise entwickelt werden: Dazu werde aus den Streckfaktoren $x_1,~ x_2, \dots,~ x_n$ ein $n$-dimensionalen Vektor $ \newcommand\a[1]{% \begin{pmatrix} a_{1#1} \\ a_{2#1} \\ \vdots \\ a_{m#1} \end{pmatrix} } \def\x{\begin{pmatrix} x_1 \\ x_2 \\ \vdots \\ x_n \end{pmatrix}} % \x =: \vec{x} $ gebildet. Matrix. Des Weiteren werden die Vektoren $ \a{1},~ \a{2},~ \dots, \a{n} $ zu einer neuen Größe $\A =: A$ zusammengefasst, die den Namen "Matrix" erhält. Bemerkung: Die Notwendigkeit des Doppelindexes in der Notation $\A$ wurde bereits zu Beginn des vorangegangenen Abschnittes klar gemacht. Nach Konvention ist der erste Index der Zeilenindex, der zweite der Spaltenindex. (Merkhilfe: "Zeile zuerst, Spalte später.") Damit wird als Kurzschreibweise $ \vec{y} =A\vec{x} =\A \x \overset{\textsf{(2)}}{=}\yR =\y =\vec{y}.$ Hiermit, also mit $A\vec{x} =\vec{y}$, ist gleichwohl das Produkt $ \text{Matrix} \times \text{Vektor} = \text{neuer Vektor} $ erklärt, das sich aus der oben genannten Linearkombination berechnet.
 

Falk-Schema.

Das bis hierhin erklärte Matrixprodukt lässt sich übersichtlich mit dem Falk-Schema notieren: $\begin{array}{c | l l} & \boxed{n-\text{dimensionaler Vektor}} \\ & \x =\vec{x} \\ \hline A=\A & \y =\yR &=A\vec{x} =\vec{y} \\ \boxed{(m \times n)-\text{Matrix}} & \boxed{m-\text{dimensionaler Vektor}} \end{array}$ Prinzip: Denkt man sich Horizontale durch die Zeilen der Matrix und bringt diese mit der Vertikalen durch den Vektor zum Schnitt, so steht an den Schnittpunkten der jeweilige Zeilenwert des Produktes, der sich wie gezeigt berechnet. Damit das Produkt bildbar ist, muss die $\text{Matrix}$ soviele Spalten haben wie der $\text{Vektor}$ Komponenten bzw. Zeilen hat; in Zeichen: $ (m \times n) \cdot (n \times 1) = (m \times 1).$
 

Weitere Definitionen.

$ % Definitionen ======================== \renewcommand\top{\mathsf{Z}} % Matrix A \newcommand\A{\begin{pmatrix} a_{11} & a_{12} & \cdots & a_{1n} \\ a_{21} & a_{22} & \cdots & a_{2n} \\ \vdots & \vdots & \ddots & \vdots \\ a_{m1} & a_{m2} & \cdots & a_{mn} \\ \end{pmatrix}} % Matrix B \newcommand\B{\begin{pmatrix} b_{11} & b_{12} & \cdots & b_{1m} \\ b_{21} & b_{22} & \cdots & b_{2m} \\ \vdots & \vdots & \ddots & \vdots \\ b_{p1} & b_{p2} & \cdots & b_{pm} \\ \end{pmatrix}} % Zeilenvektor % \z{Größe}{Zeile}{Maximale Spalte} mit Doppel-Indizes \newcommand\z[3]{(#1_{#2 1},~#1_{#2 2},\dots,~#1_{#2 #3})} % Matrix aus Zeilenvektoren mit einfachem Index % \s[transponiert]{Größe}{Maximale Zeile} \newcommand\zI[3][\top]{\left\lgroup\begin{matrix} \vec{#2}_1^{#1} \\ \vec{#2}_2^{#1} \\ \vdots \\ \vec{#2}_{#3}^{#1} \end{matrix}\right\rgroup} % Matrix aus Spaltenvektoren mit einfachem Index % \s{Größe}{Maximale Spalte} \newcommand\s[2]{\left\lgroup \begin{array}{c c c c} \\[-1em] \vec{#1}_{1}, & \vec{#1}_{2}, & \dots, & \vec{#1}_{#2} \\[-1em] \hphantom{\vec{b}_p^\top \vec{a}_1} &\hphantom{\vec{b}_p^\top \vec{a}_2 } &\hphantom{\dots} & \hphantom{\vec{b}_p^\top \vec{a}_1} \end{array}\right\rgroup} % Spaltenvektor mit Koordinaten mit einfachem Index % \sII{Größe}{Maximale Zeile} \newcommand\sII[2]{\begin{pmatrix} #1_1 \\ #1_2 \\ \vdots \\ #1_{#2} \end{pmatrix}} % Spaltenvektor mit Koordinaten mit Doppel-Index % \sI{Größe}{Spalte}{Maximale Zeile} \newcommand\sI[3]{\begin{pmatrix} #1_{1 #2} \\ #1_{2 #2} \\ \vdots \\ #1_{#3 #2} \end{pmatrix}} % Zeile im Matrixprodukt a11 x1 +a12 x2 +...+a1n xn \newcommand\AxZ[1]{% a_{#1 1} x_1 +a_{#1 2} x_2 +\dots +a_{#1 n} x_n} % Zeile im Matrixprodukt a11 x1 +a12 x2 +...+a1n xn % \BAZ{Zeile aus B}{Spalte aus A} \newcommand\BAZ[2]{% b_{#1 1}a_{1 #2} +b_{#1 2}a_{2 #2} +\dots +b_{#1 m}a_{m #2}} % 1. Koordinate: % (b_{11}a_{11} +b_{12}a_{21} +.... +b_{1m}a_{m1}) x1 % + (b_{11}a_{12} +b_{12}a_{22} +... +b_{1m}a_{m2}) x2 % ............................... % +(b_{11}a_{1n} +b_{12}a_{2n} +... +b_{1m}a_{mn}) xn % 2. Koordinate: % (b_{21}a_{11} +b_{22}a_{21} +.... +b_{2m}a_{m1}) x1 % + (b_{21}a_{12} +b_{22}a_{22} +... +b_{2m}a_{m2}) x2 % ............................... % +(b_{21}a_{1n} +b_{22}a_{2n} +... +b_{2m}a_{mn}) xn % ................ % ................ % ................ % Matrixprodukt Ax =y \newcommand\Ax{\begin{pmatrix} \AxZ{1} \\ \AxZ{2} \\ \dots \dots \dots \dots \dots \dots \dots \dots \dots \\ \AxZ{m} \end{pmatrix}} % Matrixprodukt By \newcommand\By{\begin{pmatrix} \vec{b}_1^\top \vec{y} \\ \vec{b}_2^\top \vec{y} \\ \vdots \\ \vec{b}_p^\top \vec{y} \\ \end{pmatrix}} % By ausgeschrieben \newcommand\ByI{\begin{pmatrix} \z{b}{1}{m}\Ax \\ \\ \z{b}{2}{m}\Ax \\[2.5em] \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \\[1em] \z{b}{p}{m}\Ax \\ \end{pmatrix}} % By ausgeschrieben mit Zeilen im Matrixprodukt \newcommand\ByII{\left(\begin{array}{l} ~b_{11}\, (\AxZ{1}) \\ \hspace{0.5em} +b_{12}\, (\AxZ{2}) \\ \hspace{0.5em} +\dots \\ \hspace{0.5em} +b_{1m}\, (\AxZ{m}) \\[1em] % b_{21}\, (\AxZ{1}) \\ \hspace{0.5em} +b_{22}\, (\AxZ{2}) \\ \hspace{0.5em} +\dots \\ \hspace{0.5em} +b_{2m}\, (\AxZ{m}) \\[1em] % \dots \dots \dots \dots \dots \dots \dots \dots \dots \\[1em] % b_{p1}\, (\AxZ{1}) \\ \hspace{0.5em} +b_{p2}\, (\AxZ{2}) \\ \hspace{0.5em} +\dots \\ \hspace{0.5em} +b_{pm}\, (\AxZ{m}) \\ \end{array}\right)} % By ausgeschrieben mit Zeilen im Matrixprodukt nach x'en sortiert \newcommand\ByIII{\left(\begin{array}{l} % 1. Koordinate (\BAZ{1}{1})\, x_1 \\ \hspace{0.5em} +(\BAZ{1}{2})\, x_2 \\ \hspace{0.5em} +\dots \\ \hspace{0.5em} +(\BAZ{1}{n})\, x_n \\[1em] % 2. Koordinate (\BAZ{2}{1})\, x_1 \\ \hspace{0.5em} +(\BAZ{2}{2})\, x_2 \\ \hspace{0.5em} +\dots \\ \hspace{0.5em} +(\BAZ{2}{n})\, x_n \\[1em] % ... \dots \dots \dots \dots \dots \dots \dots \dots \dots \\[1em] % p-te Koordinate (\BAZ{p}{1})\, x_1 \\ \hspace{0.5em} +(\BAZ{p}{2})\, x_2 \\ \hspace{0.5em} +\dots \\ \hspace{0.5em} +(\BAZ{p}{n})\, x_n \\[1em] \end{array}\right)} \newcommand\ByIV{\left(\begin{array}{l} % 1. Koordinate \left\lgroup \z{b}{1}{m} \sI{a}{1}{m} \right\rgroup x_1 +\left\lgroup \z{b}{1}{m} \sI{a}{2}{m} \right\rgroup x_2 +\dots +\left\lgroup \z{b}{2}{m} \sI{a}{n}{m} \right\rgroup x_n \\ % 2. Koordinate \left\lgroup \z{b}{2}{m} \sI{a}{1}{m} \right\rgroup x_1 +\left\lgroup \z{b}{2}{m} \sI{a}{2}{m} \right\rgroup x_2 +\dots +\left\lgroup \z{b}{2}{m} \sI{a}{n}{m} \right\rgroup x_n \\[1em] \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \\[1em] % p-te Koordinate \left\lgroup \z{b}{p}{m} \sI{a}{1}{m} \right\rgroup x_1 +\left\lgroup \z{b}{p}{m} \sI{a}{2}{m} \right\rgroup x_2 +\dots +\left\lgroup \z{b}{p}{m} \sI{a}{n}{m} \right\rgroup x_n \end{array}\right)} \newcommand\ByV{\left(\begin{array}{l} % 1. Koordinate \left\lgroup \vec{b}_1^\top \vec{a}_1 \right\rgroup x_1 +\left\lgroup \vec{b}_1^\top \vec{a}_2 \right\rgroup x_2 +\dots +\left\lgroup \vec{b}_1^\top \vec{a}_n \right\rgroup x_n \\[1em] % 2. Koordinate \left\lgroup \vec{b}_2^\top \vec{a}_1 \right\rgroup x_1 +\left\lgroup \vec{b}_2^\top \vec{a}_2 \right\rgroup x_2 +\dots +\left\lgroup \vec{b}_2^\top \vec{a}_n \right\rgroup x_n \\[1em] \dots \dots \dots \dots \dots \dots \dots \dots \dots \\[1em] % p-te Koordinate \left\lgroup \vec{b}_p^\top \vec{a}_1 \right\rgroup x_1 +\left\lgroup \vec{b}_p^\top \vec{a}_2 \right\rgroup x_2 +\dots +\left\lgroup \vec{b}_p^\top \vec{a}_n \right\rgroup x_n \end{array}\right)} \newcommand\ByVI{\left(\begin{array}{c c c c} % 1. Koordinate \vec{b}_1^\top \vec{a}_1 \ &\vec{b}_1^\top \vec{a}_2 &\dots &\vec{b}_1^\top \vec{a}_n \\ % 2. Koordinate \vec{b}_2^\top \vec{a}_1 &\vec{b}_2^\top \vec{a}_2 &\dots &\vec{b}_2^\top \vec{a}_n \\ \vdots & \vdots & \ddots & \vdots \\ % p-te Koordinate \vec{b}_p^\top \vec{a}_1 &\vec{b}_p^\top \vec{a}_2 &\dots &\vec{b}_p^\top \vec{a}_n \end{array}\right)} % Annotationen \def\ByIText{\longleftarrow~\textsf{Vektoren ausschreiben}} \def\ByIIText{\longleftarrow~\textsf{Skalarprodukte ausrechnen}} \def\ByIIIText{\longleftarrow~\begin{array}[t]{l} \textsf{koordinatenweise nach $x_1,x_2,\dots, x_n$ sortieren;} \\ \textsf{in folgender Zeile:} \\ \textsf{Sortierungen wieder zu Skalarprodukten zusammenfassen} \\ \downarrow \end{array}} \def\ByVText{\longleftarrow~\textsf{Kurzscheibweisen verwenden}} \def\ByVIText{\longleftarrow~\textsf{Linearkombination als Matrixprodukt ($\mathrm{Matrix}\times\mathrm{Vektor}$) umschreiben}} $ · Zeilenvektor. Um im Folgenden die Notation möglichst einfach zu halten, sei unter $ \vec{b}_i^\top =\z{b}{i}{m} $ der Zeilenvektor der $i$-ten Zeile einer Matrix $B=\B=\zI[\top]{b}{p}$ verstanden. Also zum Beispiel $\vec{b}_2^\top =\z{b}{2}{m}.$ · Spaltenvektor. Unter $\vec{a}_k =\sI{a}{k}{m}$ sei der Spaltenvektor der $k$-ten Spalte einer Matrix $A=\A= \s{a}{n}$ verstanden. Also zum Beispiel $\vec{a}_2 =\sI{a}{2}{m}.$ · Skalarprodukt. Unter der reellen Zahl $\boxed{ \vec{v}^\top \vec{w} =(v_1,~ v_2,~\dots,~ v_s) \begin{pmatrix} w_1 \\ w_2 \\ \vdots \\ w_s \end{pmatrix} =v_1 w_1 + v_2 w_2 +\dots+v_s w_s }$ sei das Skalarprodukt der Vektoren $\vec{v}^\top$ und $\vec{w}$ verstanden. Also zum Beispiel $\vec{b}_1^\top \vec{a}_2 =\z{b}{1}{m} \sI{a}{2}{m} =b_{11}a_{12} +b_{12}a_{22}+\dots+b_{1m}a_{m2}$ bezüglich oben genannter Matrizen $A$ und $B$.
 
Das bereits aufgestellte Produkt $ \text{Matrix}\times\text{Vektor}=\text{neuer Vektor} $ lässt sich mit Hilfe von Skalarprodukten ausdrücken, abermals mit dem Falk-Schema notiert: $ % Ax ausgeschrieben als Skalarprodukte \newcommand\AxI{\begin{pmatrix} \vec{a}_1^\top \vec{x} \\ \vec{a}_2^\top \vec{x} \\ \vdots \\ \vec{a}_m^\top \vec{x} \\ \end{pmatrix}} $ $ \begin{array}{c | c l l} && \boxed{n-\text{dimensionaler Vektor}} \\ & \vec{x} &=\x \\ \hline A=\A=\zI[\top]{a}{m} & \AxI &=\yR =\y &=A\vec{x} =\vec{y} \\ \boxed{(m \times n)-\text{Matrix}} && \boxed{m-\text{dimensionaler Vektor}} \end{array}$
 

Das Produkt $\text{Multiplikanden-Matrix} \times \text{Multiplikator-Matrix}.$

Wählt man die Zahlen $y_1,~ y_2,\dots, y_m \in\mathbb{R}$ aus $ \y =\Ax =A\vec{x} =\vec{y}$ als Streckfaktoren für $p$-dimensionale Vektoren $ % \newcommand\b[1]{% \begin{pmatrix} b_{1#1} \\ b_{2#1} \\ \vdots \\ b_{p#1} \end{pmatrix} } \vec{b}_1=\b{1},~ \vec{b}_2=\b{2},~ \dots,~ \vec{b}_m=\b{m},$ die in einer Anzahl $m$ vorgelegt sind, so ist die Linearkombination $ \vec{b}_1 y_1 +\vec{b}_2 y_2 +\dots +\vec{b}_m y_m =\b{1}y_1 +\b{2}y_2 +\dots +\b{m}y_m$ zu berechnen. Mit Hilfe der Matrix $B :=\B$ schreibt sich dafür wieder abkürzend $\begin{array}{l} B\vec{y} &=\B \y \\[1em] &=\B \Ax \\[1em] &=B\, (A\, \vec{x}) \end{array}$
 
Ziel. Ziel ist es nunmehr zu zeigen, dass $B\, (A\, \vec{x}) = (B\, A)\vec{x}$ gilt; wobei $B\, A$ eine neue Matrix darstellt und rechnerisch allein das zuvor erklärte Produkt $\text{Matrix}\times\text{Vektor}=\text{neuer Vektor}$ benötigt wird.

 
$% Definitionen ======================== % %\renewcommand\top{\textsf{S}} % % Matrix A \newcommand\A{\begin{pmatrix} a_{11} & a_{12} & \cdots & a_{1n} \\ a_{21} & a_{22} & \cdots & a_{2n} \\ \vdots & \vdots & \ddots & \vdots \\ a_{m1} & a_{m2} & \cdots & a_{mn} \\ \end{pmatrix}} % Matrix B \newcommand\B{\begin{pmatrix} b_{11} & b_{12} & \cdots & b_{1m} \\ b_{21} & b_{22} & \cdots & b_{2m} \\ \vdots & \vdots & \ddots & \vdots \\ b_{p1} & b_{p2} & \cdots & b_{pm} \\ \end{pmatrix}} % Zeilenvektor % \z{Größe}{Zeile}{Maximale Spalte} mit Doppel-Indizes \newcommand\z[3]{(#1_{#2 1},~#1_{#2 2},\dots,~#1_{#2 #3})} % Matrix aus Zeilenvektoren mit einfachem Index % \zI[transponiert]{Größe}{Maximale Zeile} \newcommand\zI[3][\top]{\left\lgroup\begin{matrix} \vec{#2}_1^{#1} \\ \vec{#2}_2^{#1} \\ \vdots \\ \vec{#2}_{#3}^{#1} \end{matrix}\right\rgroup} % Matrix aus Spaltenvektoren mit einfachem Index % \s{Größe}{Maximale Spalte} \newcommand\s[2]{\left\lgroup \begin{array}{c c c c} \\[-1em] \vec{#1}_{1}, & \vec{#1}_{2}, & \dots, & \vec{#1}_{#2} \\[-1em] \hphantom{\vec{b}_p^\top \vec{a}_1} &\hphantom{\vec{b}_p^\top \vec{a}_2 } &\hphantom{\dots} & \hphantom{\vec{b}_p^\top \vec{a}_1} \end{array}\right\rgroup} % Spaltenvektor mit Koordinaten mit einfachem Index % \sII{Größe}{Maximale Zeile} \newcommand\sII[2]{\begin{pmatrix} #1_1 \\ #1_2 \\ \vdots \\ #1_{#2} \end{pmatrix}} % Spaltenvektor mit Koordinaten mit Doppel-Index % \sI{Größe}{Spalte}{Maximale Zeile} \newcommand\sI[3]{\begin{pmatrix} #1_{1 #2} \\ #1_{2 #2} \\ \vdots \\ #1_{#3 #2} \end{pmatrix}} % Zeile im Matrixprodukt a11 x1 +a12 x2 +...+a1n xn \newcommand\AxZ[1]{% a_{#1 1} x_1 +a_{#1 2} x_2 +\dots +a_{#1 n} x_n} % Zeile im Matrixprodukt a11 x1 +a12 x2 +...+a1n xn % \BAZ{Zeile aus B}{Spalte aus A} \newcommand\BAZ[2]{% b_{#1 1}a_{1 #2} +b_{#1 2}a_{2 #2} +\dots +b_{#1 m}a_{m #2}} % 1. Koordinate: % (b_{11}a_{11} +b_{12}a_{21} +.... +b_{1m}a_{m1}) x1 % + (b_{11}a_{12} +b_{12}a_{22} +... +b_{1m}a_{m2}) x2 % ............................... % +(b_{11}a_{1n} +b_{12}a_{2n} +... +b_{1m}a_{mn}) xn % 2. Koordinate: % (b_{21}a_{11} +b_{22}a_{21} +.... +b_{2m}a_{m1}) x1 % + (b_{21}a_{12} +b_{22}a_{22} +... +b_{2m}a_{m2}) x2 % ............................... % +(b_{21}a_{1n} +b_{22}a_{2n} +... +b_{2m}a_{mn}) xn % ................ % ................ % ................ % Matrixprodukt Ax =y \newcommand\Ax{\begin{pmatrix} \AxZ{1} \\ \AxZ{2} \\ \dots \dots \dots \dots \dots \dots \dots \dots \dots \\ \AxZ{m} \end{pmatrix}} % Matrixprodukt By ausgeschrieben als Skalarprodukte \newcommand\By{\begin{pmatrix} \vec{b}_1^\top \vec{y} \\ \vec{b}_2^\top \vec{y} \\ \vdots \\ \vec{b}_p^\top \vec{y} \\ \end{pmatrix}} % By ausgeschrieben \newcommand\ByI{\begin{pmatrix} \z{b}{1}{m}\Ax \\ \\ \z{b}{2}{m}\Ax \\[2.5em] \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \\[1em] \z{b}{p}{m}\Ax \\ \end{pmatrix}} % By ausgeschrieben mit Zeilen im Matrixprodukt \newcommand\ByII{\left(\begin{array}{l} ~b_{11}\, (\AxZ{1}) \\ \hspace{0.5em} +b_{12}\, (\AxZ{2}) \\ \hspace{0.5em} +\dots \\ \hspace{0.5em} +b_{1m}\, (\AxZ{m}) \\[1em] % b_{21}\, (\AxZ{1}) \\ \hspace{0.5em} +b_{22}\, (\AxZ{2}) \\ \hspace{0.5em} +\dots \\ \hspace{0.5em} +b_{2m}\, (\AxZ{m}) \\[1em] % \dots \dots \dots \dots \dots \dots \dots \dots \dots \\[1em] % b_{p1}\, (\AxZ{1}) \\ \hspace{0.5em} +b_{p2}\, (\AxZ{2}) \\ \hspace{0.5em} +\dots \\ \hspace{0.5em} +b_{pm}\, (\AxZ{m}) \\ \end{array}\right)} % By ausgeschrieben mit Zeilen im Matrixprodukt nach x'en sortiert \newcommand\ByIII{\left(\begin{array}{l} % 1. Koordinate (\BAZ{1}{1})\, x_1 \\ \hspace{0.5em} +(\BAZ{1}{2})\, x_2 \\ \hspace{0.5em} +\dots \\ \hspace{0.5em} +(\BAZ{1}{n})\, x_n \\[1em] % 2. Koordinate (\BAZ{2}{1})\, x_1 \\ \hspace{0.5em} +(\BAZ{2}{2})\, x_2 \\ \hspace{0.5em} +\dots \\ \hspace{0.5em} +(\BAZ{2}{n})\, x_n \\[1em] % ... \dots \dots \dots \dots \dots \dots \dots \dots \dots \\[1em] % p-te Koordinate (\BAZ{p}{1})\, x_1 \\ \hspace{0.5em} +(\BAZ{p}{2})\, x_2 \\ \hspace{0.5em} +\dots \\ \hspace{0.5em} +(\BAZ{p}{n})\, x_n \\[1em] \end{array}\right)} \newcommand\ByIV{\left(\begin{array}{l} % 1. Koordinate \left\lgroup \z{b}{1}{m} \sI{a}{1}{m} \right\rgroup x_1 +\left\lgroup \z{b}{1}{m} \sI{a}{2}{m} \right\rgroup x_2 +\dots +\left\lgroup \z{b}{1}{m} \sI{a}{n}{m} \right\rgroup x_n \\ % 2. Koordinate \left\lgroup \z{b}{2}{m} \sI{a}{1}{m} \right\rgroup x_1 +\left\lgroup \z{b}{2}{m} \sI{a}{2}{m} \right\rgroup x_2 +\dots +\left\lgroup \z{b}{2}{m} \sI{a}{n}{m} \right\rgroup x_n \\[1em] \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \\[1em] % p-te Koordinate \left\lgroup \z{b}{p}{m} \sI{a}{1}{m} \right\rgroup x_1 +\left\lgroup \z{b}{p}{m} \sI{a}{2}{m} \right\rgroup x_2 +\dots +\left\lgroup \z{b}{p}{m} \sI{a}{n}{m} \right\rgroup x_n \end{array}\right)} \newcommand\ByV{\left(\begin{array}{l} % 1. Koordinate \left\lgroup \vec{b}_1^\top \vec{a}_1 \right\rgroup x_1 +\left\lgroup \vec{b}_1^\top \vec{a}_2 \right\rgroup x_2 +\dots +\left\lgroup \vec{b}_1^\top \vec{a}_n \right\rgroup x_n \\[1em] % 2. Koordinate \left\lgroup \vec{b}_2^\top \vec{a}_1 \right\rgroup x_1 +\left\lgroup \vec{b}_2^\top \vec{a}_2 \right\rgroup x_2 +\dots +\left\lgroup \vec{b}_2^\top \vec{a}_n \right\rgroup x_n \\[1em] \dots \dots \dots \dots \dots \dots \dots \dots \dots \\[1em] % p-te Koordinate \left\lgroup \vec{b}_p^\top \vec{a}_1 \right\rgroup x_1 +\left\lgroup \vec{b}_p^\top \vec{a}_2 \right\rgroup x_2 +\dots +\left\lgroup \vec{b}_p^\top \vec{a}_n \right\rgroup x_n \end{array}\right)} \newcommand\ByVI{\begin{pmatrix} \vec{b}_1^\top \vec{a}_1 \\[1em] \vec{b}_2^\top \vec{a}_1 \\[1em] \vdots \\[1em] \vec{b}_p^\top \vec{a}_1 \\[1em] \end{pmatrix} x_1 +\begin{pmatrix} \vec{b}_1^\top \vec{a}_2 \\[1em] \vec{b}_2^\top \vec{a}_2 \\[1em] \vdots \\[1em] \vec{b}_p^\top \vec{a}_2 \\[1em] \end{pmatrix} x_2 +~\dots~ +\begin{pmatrix} \vec{b}_1^\top \vec{a}_n \\[1em] \vec{b}_2^\top \vec{a}_n \\[1em] \vdots \\[1em] \vec{b}_p^\top \vec{a}_n \\[1em] \end{pmatrix} x_n} \newcommand\ByVII{\left(\begin{array}{c c c c} % 1. Koordinate \vec{b}_1^\top \vec{a}_1 \ &\vec{b}_1^\top \vec{a}_2 &\dots &\vec{b}_1^\top \vec{a}_n \\ % 2. Koordinate \vec{b}_2^\top \vec{a}_1 &\vec{b}_2^\top \vec{a}_2 &\dots &\vec{b}_2^\top \vec{a}_n \\ \vdots & \vdots & \ddots & \vdots \\ % p-te Koordinate \vec{b}_p^\top \vec{a}_1 &\vec{b}_p^\top \vec{a}_2 &\dots &\vec{b}_p^\top \vec{a}_n \end{array}\right)} % Annotationen \def\ByIText{\longleftarrow~\textsf{Vektoren ausschreiben}} \def\ByIIText{\longleftarrow~\textsf{Skalarprodukte ausrechnen}} \def\ByIIIText{\longleftarrow~\begin{array}[t]{l} \textsf{koordinatenweise nach $x_1,x_2,\dots, x_n$ sortieren;} \\ \textsf{in folgender Zeile:} \\ \textsf{Sortierungen wieder zu} \\ \textsf{Skalarprodukten zusammenfassen} \\ \downarrow \end{array}} \def\ByVText{\longleftarrow~\begin{array}[t]{l} \textsf{Kurzscheibweisen mit} \\ \textsf{Zeilen- und Spaltenvektoren für} \\ \textsf{die Skalarprodukte verwenden} \end{array}} \def\ByVIText{\longleftarrow~\begin{array}[t]{l} \textsf{Linearkombination in üblicher} \\ \textsf{Notation angeben} \end{array}} \def\ByVIIText{\longleftarrow~\begin{array}[t]{l} \textsf{Linearkombination als Matrix-} \\ \textsf{produkt ($\mathrm{Matrix}\times\mathrm{Vektor}$)} \\ \textsf{umschreiben} \end{array}} $ Zunächst wird die Matrix $B$ mit Hilfe von Zeilenvektoren abkürzend umgeschrieben; dann ist im Falk-Schema notiert: $\begin{array}{l | c l} & \vec{y} &=\sII{y}{m} =\Ax =A\vec{x} \\ \hline B =\B=\zI[\top]{b}{p} & \By &=B\vec{y} =B(A\vec{x}) \end{array}$ Es ist also zu berechnen $\begin{array}{l l} B\vec{y} &=B(A\vec{x})= \By \\ &= \ByI \ByIText \\ \\ &= \ByII \ByIIText \\ \\ &= \ByIII \ByIIIText \\ \\ &={\small \ByIV} \\ \\ &= \ByV \ByVText\\ \\[1em] &= \ByVI \ByVIText \\ \\[1em] &= \ByVII \sII{x}{n} \ByVIIText \\ \\[1em] &=\underline{\underline{ ~(BA)\vec{x}~ }} \end{array}$
 

Das allgemeine Matrixprodukt.

Der vorletzten Zeile in vorangehender Rechnung konnte dabei eine Berechnungsvorschrift für das Matrixprodukt $B\, A$ entlesen werden; wieder im Falk-Schema notiert: $\begin{array}{c | c l} & \boxed{(m\times n)\textsf{-Matrix}} \\ & \s{a}{n} &=\A =A \\ \hline B =\B=\zI[\top]{b}{p} & \ByVII &=BA \rule{0pt}{14mm} \\[1em] \boxed{(p\times m)\textsf{-Matrix}} & \boxed{(p\times n)\textsf{-Matrix}} \end{array}$ Falk-Schema: Denkt man sich Horizontale durch die Zeilen der $\text{Multiplikanden-Matrix} ~B$ und bringt diese mit der Vertikalen durch die Spalten der $\text{Multiplikator-Matrix} ~A$ zum Schnitt, so steht an den Schnittpunkten der jeweilige Wert des Matrixelements (berechnet als Skalarprodukt) der Produktmatrix $B\, A$. Es wurde also aus der Linearkombination von Vektoren, also dem Matrixprodukt $ \text{Matrix} \times \text{Vektor} =\text{neuer Vektor},$ eine Verallgemeinerung $ \text{Multiplikanden-Matrix} \times \text{Multiplikator-Matrix} =\text{neue Matrix}$ hergeleitet. Damit Multiplizierbarkeit gegeben ist, muss die $\text{Multiplikanden-Matrix}$ so viele Spalten enthalten, wie die $\text{Multiplikator-Matrix}$ Zeilen besitzt; in Zeichen: $ (p \times m) \cdot (m \times n) = (p \times n).$ Gleichwohl ist damit gezeigt, dass für Matrixprodukte das Kommutativgesetz (Vertauschungsregel) nicht gilt: im allgemeinen ist $B\, A \neq A\,B.$
 

Beispiel zur Veranschaulichung einer Matrixmultiplikation.

\def\lI{1.5} \def\lII{0.5} \def\lIII{3} \def\lIV{-1} \scalebox{0.625}{ \begin{tikzpicture}[ gridded, scale=1, >=latex, every path/.style={thick, ->}, ] %A \coordinate[label=] (S) at (0,0); \coordinate[label=] (A) at (2,2); \coordinate[label=] (As) at ($\lI*(A)$); \draw[->] (S) -- (A) node[midway, left=2.5mm]{$\dbinom{2}{2}$}; \draw[->, blue,] ([xshift=1cm]S) -- ([xshift=1cm]As) node[pos=0.3, right]{$\lI\, \dbinom{2}{2}$}; % Rechnung \node[red, fill=lightgray, anchor=north west, yshift=-9.75cm] at (S){ $\begin{array}{l l} \begin{pmatrix} 2 & 4 & 0 & 3 \\ 2 & 0 & 1 & 1 \end{pmatrix} \begin{pmatrix} \begin{array}{r}1.5 \\ 0.5 \\ 3 \\ -1 \end{array}\end{pmatrix} &=\begin{pmatrix} 2\cdot 1.5 + 4\cdot 0.5 +0\cdot3 +3\cdot(-1) \\[1em] 2\cdot 1.5 + 0\cdot 0.5 +1\cdot3 +1\cdot(-1) \end{pmatrix} =\begin{pmatrix} 2 \\5 \end{pmatrix} \\ \\ &=\lI\, \dbinom{2}{2} +\lII\, \dbinom{4}{0} +\lIII\, \dbinom{0}{1} -\dbinom{3}{1} =\begin{pmatrix} 2 \\5 \end{pmatrix} \end{array}$}; %B \begin{scope}[shift={(4,0)}] \coordinate[label=] (S) at (0,0); \coordinate[label=] (B) at (4,0); \coordinate[label=] (Bs) at ($\lII*(B)$); \draw[->] (S) -- (B) node[pos=0.75, sloped, above]{$\dbinom{4}{0}$}; \draw[->, blue,] ([shift={(0,1)}]S) -- ([shift={(0,1)}]Bs) node[midway, sloped, above]{$\lII\, \dbinom{4}{0}$}; \end{scope} %C \begin{scope}[shift={(9,0)}] \coordinate[label=] (S) at (0,0); \coordinate[label=] (C) at (0,1); \coordinate[label=] (Cs) at ($\lIII*(C)$); \draw[->] (S) -- (C) node[midway, left]{$\dbinom{0}{1}$}; \draw[->, blue,] ([shift={(1,0)}]S) -- ([shift={(1,0)}]Cs) node[midway, right]{$\lIII\, \dbinom{0}{1}$}; \end{scope} %D \begin{scope}[shift={(12,0)}] \coordinate[label=] (S) at (0,0); \coordinate[label=] (D) at (3,1); \coordinate[label=] (Ds) at ($\lIV*(D)$); \draw[->] (S) -- (D) node[midway, below]{$\dbinom{3}{1}$}; \draw[->, blue,] ([shift={(3,2)}]S) -- ([shift={(3,2)}]Ds) node[midway, above]{$-\dbinom{3}{1}$}; \end{scope} \begin{scope}[shift={(5,-8)}] \coordinate[label=] (S) at (0,0); \coordinate[label=] (A) at (2,2); \coordinate[label=] (As) at ($\lI*(A)$); \coordinate[label=] (B) at (4,0); \coordinate[label=] (Bs) at ($\lII*(B)$); \coordinate[label=] (C) at (0,1); \coordinate[label=] (Cs) at ($\lIII*(C)$); \coordinate[label=] (D) at (3,1); \coordinate[label=] (Ds) at ($\lIV*(D)$); \draw[blue] (S) coordinate (X0) -- (As) coordinate (X1) node[midway, sloped, below]{$\lI\,\dbinom{2}{2}$}; \draw[blue] (X1) -- +(Bs) coordinate (X2) node[midway, sloped, above]{$\lII\, \dbinom{4}{0}$}; \draw[blue] (X2) -- +(Cs) coordinate (X3) node[midway, right]{$\lIII\,\dbinom{0}{1}$}; \draw[blue] (X3) -- +(Ds) coordinate (X4) node[midway, sloped, above]{$-\dbinom{3}{1}$}; \draw[very thick, red] (S) -- (X4) node[align=left, midway, right, anchor=east]{ $ \begin{pmatrix} 2 & 4 & 0 & 3 \\ 2 & 0 & 1 & 1 \end{pmatrix} \begin{pmatrix} \begin{array}{r}1.5 \\ 0.5 \\ 3 \\ -1 \end{array}\end{pmatrix} =\begin{pmatrix} 2 \\ 5 \end{pmatrix} $}; \foreach \k in {0,...,4} \draw[fill=black!1] (X\k) circle(2.0pt); \end{scope} \end{tikzpicture} }
 

Quellen.

Matrix in "Mathematische Formelsammlung - Begriffe und Sätze (Diesterweg 1989)" Matrix auf wikipedia: de.wikipedia.org/wiki/Matrix_(Mathematik) Falk-Schema: de.wikipedia.org/wiki/Falkschema
 
Danke @thureduehrsen für die Durchsicht.
 


Dieser Artikel kommt von Matroids Matheplanet
https://matheplanet.de

Die Url für diesen Artikel ist:
https://matheplanet.de/default3.html?article=2001