【行列】特異値分解~概要と例題~

行列

 本記事では、「行列の特異値分解」を扱う。

 簡単に言うと特異値分解とは、正方行列の対角化を一般の行列に拡張したものと考えてよい。

 正方行列の対角化は下記を参照。

 

 厳密な議論は教科書に任せて、本記事では特異値分解の計算に必要な事項を中心にまとめる。

広告

概説

 任意の\(m\times n\)行列\(\mathsf{A}\)を考える。ただし\(m \neq n\)であり、\(\mathsf{A}\)はゼロ行列ではない。

\begin{align}
\mathsf{A}=\begin{pmatrix}
a_{1,1}&a_{1,2}&\cdots&a_{1,n}\\
a_{2,1}&a_{2,2}&\cdots&a_{2,n}\\
\vdots & \vdots & \ddots &\vdots \\
a_{m,1}&a_{m,2}&\cdots&a_{m,n}
\end{pmatrix}\tag{1}
\end{align}

 ここで、ともにゼロベクトルでない\(m\)次元ベクトル\(\vec{u}=(u_{1},u_{2},\cdots ,u_{m})\)、\(n\)次元ベクトル\(\vec{v}=(v_{1},v_{2},\cdots ,v_{n})\)に対して

\begin{gather}
\mathsf{A}\vec{v}=\sigma\vec{u} \quad \Rightarrow \quad
\begin{pmatrix}
a_{1,1}&a_{1,2}&\cdots&a_{1,n}\\
a_{2,1}&a_{2,2}&\cdots&a_{2,n}\\
\vdots & \vdots & \ddots &\vdots \\
a_{m,1}&a_{m,2}&\cdots&a_{m,n}
\end{pmatrix}\begin{pmatrix}v_{1}\\ v_{2}\\ \vdots\\ v_{n}\end{pmatrix}
=\sigma\begin{pmatrix}u_{1}\\ u_{2}\\ \vdots \\ u_{m}\end{pmatrix} \label{tokui1} \tag{2}
\end{gather}

および

\begin{gather}
\mathsf{A}^{\text{T}}\vec{u}=\sigma\vec{v} \quad \Rightarrow \quad
\begin{pmatrix}
a_{1,1}&a_{2,1}&\cdots&a_{m,1}\\
a_{1,2}&a_{2,2}&\cdots&a_{m,2}\\
\vdots & \vdots & \ddots &\vdots \\
a_{1,n}&a_{2,n}&\cdots&a_{m,n}
\end{pmatrix}\begin{pmatrix}u_{1}\\ u_{2}\\ \vdots\\ u_{m}\end{pmatrix}
=\sigma\begin{pmatrix}v_{1}\\ v_{2}\\ \vdots \\ v_{n}\end{pmatrix} \label{tokui2} \tag{3}
\end{gather}

を満たす\(\sigma>0\)が存在するとき、\(\sigma\)を特異値、\(\vec{u}\)を左特異ベクトル、\(\vec{v}\)を右特異ベクトルと呼ぶ。

  

 

 特異値と特異ベクトルは以下のようにして求める。

 まず(\ref{tokui1})の両辺に左から\(\mathsf{A}^{\text{T}}\)を作用させると、(\ref{tokui2})より

\begin{align}
&\mathsf{A}^{\text{T}}\mathsf{A}\vec{v}=\mathsf{A}^{\text{T}}\sigma\vec{u}=\sigma(\mathsf{A}^{\text{T}}\vec{u})=\sigma(\sigma\vec{v})=\sigma^{2}\vec{v} &\therefore \mathsf{A}^{\text{T}}\mathsf{A}\vec{v}=\sigma^{2}\vec{v} \label{ata} \tag{4}
\end{align}

となる。

 また(\ref{tokui2})の両辺に左から\(\mathsf{A}\)を作用させると、(\ref{tokui1})より

\begin{align}
&\mathsf{A}\mathsf{A}^{\text{T}}\vec{u}=\mathsf{A}\sigma\vec{v}=\sigma(\mathsf{A}\vec{v})=\sigma(\sigma\vec{u})=\sigma^{2}\vec{u} &\therefore \mathsf{A}\mathsf{A}^{\text{T}}\vec{u}=\sigma^{2}\vec{u} \label{aat} \tag{5}
\end{align}

となる。

  \(\mathsf{A}^{\text{T}}\mathsf{A}\)は\(n\)次元の正方行列、\(\mathsf{A}\mathsf{A}^{\text{T}}\)は\(m\)次元の正方行列である。

 よってこの2つの正方行列の固有値と固有ベクトルを求めれば、特異値と特異ベクトルを求めることが出来る。

 特に行列\(\mathsf{A}\)が実行列である場合、\(\mathsf{A}^{\text{T}}\mathsf{A}\)と\(\mathsf{A}\mathsf{A}^{\text{T}}\)はともに実対称行列となる。

 よって固有値は全て実数であり、固有ベクトルは正規直交基底をなすように取ることができる。

 

 特異値は\(n\)または\(m\)の少ない方の数だけ得られる。

 すなわち特異値の個数を\(r\)とすると、\(r=\text{min}\{m,n\}\)となる。

 よって特異値に関しては\(\mathsf{A}^{\text{T}}\mathsf{A}\)または\(\mathsf{A}\mathsf{A}^{\text{T}}\)のうち、次数が少ない方の正方行列を選んで固有値を求めるのが手っ取り早い。

 

 こうして得られた特異値を大きい順、すなわち\(\sigma_{1}\geq\sigma_{2}\geq…\geq\sigma_{r}>0\)と対角に並べ、それ以外の成分を0とした\(m\times n\)行列\(\Sigma\)

\begin{align}
\begin{matrix}
\Sigma=\begin{pmatrix}
\sigma_{1}&0&\cdots &0&\cdots&0 \\
0&\sigma_{2}&\cdots &0&\cdots&0\\
\vdots & \vdots & \ddots & \vdots&\ddots&\vdots\\
0&0&\cdots&\sigma_{m}&\cdots&0
\end{pmatrix}&\text{or}&
\Sigma=\begin{pmatrix}
\sigma_{1}&0&\cdots &0 \\
0&\sigma_{2}&\cdots &0\\
\vdots & \vdots & \ddots & \vdots\\
0&0&\cdots&\sigma_{n}\\
\vdots & \vdots & \ddots & \vdots\\
0&0&\cdots&0
\end{pmatrix}\\ \\
(m<nの場合)&&(m>nの場合)
\end{matrix}\label{sigmamat}\tag{6}
\end{align}

左特異ベクトルを横に\(m\)個並べた\(m\)次元の正方行列\(\mathsf{U}\)

\begin{align}
\mathsf{U}=(\vec{u}^{(1)},\vec{u}^{(2)},\cdots,\vec{u}^{(m)})=\begin{pmatrix}
u_{1}^{(1)}&u_{1}^{(2)}&\cdots&u_{1}^{(m)}\\
u_{2}^{(1)}&u_{2}^{(2)}&\cdots&u_{2}^{(m)}\\
\vdots & \vdots & \ddots & \vdots\\
u_{m}^{(1)}&u_{m}^{(2)}&\cdots&u_{m}^{(m)}
\end{pmatrix}\label{Umat}\tag{7}
\end{align}

右特異ベクトルを横に\(n\)個並べた\(n\)次元の正方行列\(\mathsf{V}\)

\begin{align}
\mathsf{V}=(\vec{v}^{(1)},\vec{v}^{(2)},\cdots,\vec{v}^{(n)})=\begin{pmatrix}
v_{1}^{(1)}&v_{1}^{(2)}&\cdots&v_{1}^{(n)}\\
v_{2}^{(1)}&v_{2}^{(2)}&\cdots&v_{2}^{(n)}\\
\vdots & \vdots & \ddots & \vdots\\
v_{n}^{(1)}&v_{n}^{(2)}&\cdots&v_{n}^{(n)}
\end{pmatrix}\label{Vmat}\tag{8}
\end{align}

を用いると、元の行列\(\mathsf{A}\)は

\begin{align}
\mathsf{A}=\mathsf{U}\Sigma\mathsf{V}^{\text{T}}\label{tokuichilast}\tag{9}
\end{align}

となる。

 これが特異値分解である。

 ただし、\(\mathsf{V}^{\text{T}}\)は\(\mathsf{V}\)の転置行列である。

 

 特異値分解において、特異値を対角に大きい順に並べた\(\Sigma\)は一意に定まる(答えが一つに決まる)が、特異ベクトルを並べた\(\mathsf{U}\)と\(\mathsf{V}\)は一意に定まらない(答えが一つとは限らない)。

例題

 ここで特異値分解の例題を解いてみよう。

例題

 次の2つの行列\(\mathsf{A},\mathsf{B}\)を特異値分解せよ。

\begin{align}
\mathsf{A}=\begin{pmatrix}
1&-1&0\\
-1&0&1
\end{pmatrix},\quad
\mathsf{B}=\begin{pmatrix}
-2&0\\
0&1\\
0&-1
\end{pmatrix}\tag{10}
\end{align}

解説

 問題を通して単位行列を\(\mathsf{I}\)とする。

\(\mathsf{A}\)について

 まず特異値を求める。

 \(\mathsf{A}\)では\(\mathsf{A}^{\text{T}}\mathsf{A}\)が3次元の正方行列、\(\mathsf{A}\mathsf{A}^{\text{T}}\)が2次元の正方行列となる。

 よって\(\mathsf{A}\mathsf{A}^{\text{T}}\)の固有値を求める方が手っ取り早い。

\begin{align}
\mathsf{A}\mathsf{A}^{\text{T}}=\begin{pmatrix}2&-1\\-1&2\end{pmatrix}\tag{11}
\end{align}

より、固有値を\(\lambda\)とすると特性方程式より

\begin{gather}
\text{det}[\mathsf{A}\mathsf{A}^{\text{T}}-\lambda\mathsf{I}]=0 \\
\text{det}\left[\begin{pmatrix}2-\lambda&-1 \\ -1&2-\lambda\end{pmatrix}\right]=0 \\
(2-\lambda)^{2}-1=0 \\
\therefore \lambda=1,3\tag{12}
\end{gather}

となるため、特異値は\(\sigma=1,\sqrt{3}\)となる。

 

 続いて(\ref{aat})を利用して、特異値ごとに左特異ベクトルを求める。

 \(\mathsf{A}\mathsf{A}^{\text{T}}\)は実対称行列であるため、固有ベクトル(左特異ベクトル)は正規直交基底をなすように取ることができ、

\begin{align}
\begin{matrix}
\mathsf{A}\mathsf{A}^{\text{T}}\vec{u}^{(1)}=3\vec{u}^{(1)}&&\mathsf{A}\mathsf{A}^{\text{T}}\vec{u}^{(2)}=\vec{u}^{(2)}\\
\begin{pmatrix}2&-1\\-1&2\end{pmatrix}\begin{pmatrix}u_{1}^{(1)}\\u_{2}^{(1)}\end{pmatrix}=3\begin{pmatrix}u_{1}^{(1)}\\u_{2}^{(1)}\end{pmatrix}&&
\begin{pmatrix}2&-1\\-1&2\end{pmatrix}\begin{pmatrix}u_{1}^{(2)}\\u_{2}^{(2)}\end{pmatrix}=\begin{pmatrix}u_{1}^{(2)}\\u_{2}^{(2)}\end{pmatrix}\\ \\
\Downarrow && \Downarrow\\ \\
u_{1}^{(1)}+u_{2}^{(1)}=0 && u_{1}^{(2)}-u_{2}^{(2)}=0 \\ \\
\end{matrix}\tag{13}
\end{align}

より、左特異ベクトルは

\begin{align}
\vec{u}^{(1)}=\displaystyle{\frac{1}{\sqrt{2}}}\begin{pmatrix}1\\-1\end{pmatrix},\quad\vec{u}^{(2)}=\displaystyle{\frac{1}{\sqrt{2}}}\begin{pmatrix}1\\1\end{pmatrix} \label{ahidari}\tag{14}
\end{align}

と求められる。

 

 続いて右特異ベクトルを求める。

 まず得られた特異値と左特異ベクトルを利用すれば、(\ref{tokui2})を変形して

\begin{align}
\vec{v}^{(1)}&=\frac{1}{\sqrt{3}}\mathsf{A}^{\text{T}}\vec{u}^{(1)} & \vec{v}^{(2)}&=\mathsf{A}^{\text{T}}\vec{u}^{(2)} \\
&=\frac{1}{\sqrt{3}}\begin{pmatrix}1&-1\\-1&0\\0&1\end{pmatrix}\frac{1}{\sqrt{2}}\begin{pmatrix}1\\-1\end{pmatrix}
& &=\begin{pmatrix}1&-1\\-1&0\\0&1\end{pmatrix}\frac{1}{\sqrt{2}}\begin{pmatrix}1\\1\end{pmatrix} \\
&=\frac{1}{\sqrt{6}}\begin{pmatrix}2\\-1\\-1\end{pmatrix}&
&=\frac{1}{\sqrt{2}}\begin{pmatrix}0\\-1\\1\end{pmatrix}\tag{15}
\end{align}

と2つの右特異ベクトルが得られる。

 ここで、右特異ベクトルが\(\mathsf{A}^{\text{T}}\mathsf{A}\)の固有ベクトルであることを思い出すと、\(\mathsf{A}^{\text{T}}\mathsf{A}\)も実対称行列であるため、右特異ベクトルも正規交基底をなすように取ることができる。

 現に、\(\vec{v}^{(1)}\)と\(\vec{v}^{(1)}\)は互いに直交する単位ベクトルである。

 よってもう1つの右特異ベクトル\(\vec{v}^{(3)}\)は、右特異ベクトルも正規直交基底をなすことを利用して

\begin{gather}
\begin{cases}
\langle\vec{v}^{(1)},\vec{v}^{(3)}\rangle=0 \\
\langle\vec{v}^{(2)},\vec{v}^{(3)}\rangle=0 \\
|\vec{v}^{(3)}|=1
\end{cases} \\
\Downarrow \\
\begin{cases}
2v_{1}^{(3)}-v_{2}^{(3)}-v_{3}^{(3)}=0 \\
-v_{2}^{(3)}+v_{3}^{(3)}=0 \\
\sqrt{\left\{v_{1}^{(3)}\right\}^{2}+\left\{v_{2}^{(3)}\right\}^{2}+\left\{v_{3}^{(3)}\right\}^{2}}=1
\end{cases} \\
\Downarrow \\
\begin{cases}
v_{1}^{(3)}=v_{2}^{(3)}=v_{3}^{(3)} \\
\left\{v_{1}^{(3)}\right\}^{2}+\left\{v_{2}^{(3)}\right\}^{2}+\left\{v_{3}^{(3)}\right\}^{2}=1
\end{cases} \\
\Downarrow \\
\vec{v}^{(3)}=\frac{1}{\sqrt{3}}\begin{pmatrix}1\\1\\1\end{pmatrix}\tag{16}
\end{gather}

と求められる。

 

 以上より、

\begin{gather}
\Sigma=\begin{pmatrix}\sqrt{3}&0&0\\0&1&0\end{pmatrix}\tag{17}\\
\mathsf{U}=(\vec{u}^{(1)},\vec{u}^{(2)})=\begin{pmatrix}
\frac{1}{\sqrt{2}}&\frac{1}{\sqrt{2}}\\
-\frac{1}{\sqrt{2}}&\frac{1}{\sqrt{2}}\end{pmatrix}\tag{18}\\
\mathsf{V}=(\vec{v}^{(1)},\vec{v}^{(2)},\vec{v}^{(3)})=\begin{pmatrix}
\frac{2}{\sqrt{6}}& 0&\frac{1}{\sqrt{3}}\\
-\frac{1}{\sqrt{6}}& -\frac{1}{\sqrt{2}}&\frac{1}{\sqrt{3}}\\
-\frac{1}{\sqrt{6}} &\frac{1}{\sqrt{2}}&\frac{1}{\sqrt{3}}
\end{pmatrix}\tag{19}
\end{gather}

となるため、(\ref{tokuichilast})より

\begin{gather}
\mathsf{A}=\mathsf{U}\Sigma\mathsf{V}^{\text{T}} \\
\boxed{\begin{pmatrix}
1&-1&0\\
-1&0&1
\end{pmatrix}=\begin{pmatrix}
\frac{1}{\sqrt{2}}&\frac{1}{\sqrt{2}}\\
-\frac{1}{\sqrt{2}}&\frac{1}{\sqrt{2}}
\end{pmatrix}\begin{pmatrix}\sqrt{3}&0&0\\0&1&0\end{pmatrix}
\begin{pmatrix}
\frac{2}{\sqrt{6}}&-\frac{1}{\sqrt{6}}&-\frac{1}{\sqrt{6}}\\
0&-\frac{1}{\sqrt{2}}&\frac{1}{\sqrt{2}}\\
\frac{1}{\sqrt{3}}&\frac{1}{\sqrt{3}}&\frac{1}{\sqrt{3}}
\end{pmatrix}}\label{atokuikai}\tag{20}
\end{gather}

と特異値分解できる。

 実際に(\ref{atokuikai})の右辺を計算すると、

\begin{align}
&\begin{pmatrix}
\frac{1}{\sqrt{2}}&\frac{1}{\sqrt{2}}\\
-\frac{1}{\sqrt{2}}&\frac{1}{\sqrt{2}}
\end{pmatrix}
\begin{pmatrix}
\sqrt{3}&0&0\\
0&1&0
\end{pmatrix}
\begin{pmatrix}
\frac{2}{\sqrt{6}}&-\frac{1}{\sqrt{6}}&-\frac{1}{\sqrt{6}}\\
0&-\frac{1}{\sqrt{2}}&\frac{1}{\sqrt{2}}\\
\frac{1}{\sqrt{3}}&\frac{1}{\sqrt{3}}&\frac{1}{\sqrt{3}}
\end{pmatrix} \\
=&\begin{pmatrix}
\frac{1}{\sqrt{2}}&\frac{1}{\sqrt{2}}\\
-\frac{1}{\sqrt{2}}&\frac{1}{\sqrt{2}}\end{pmatrix}
\begin{pmatrix}
\sqrt{2}&-\frac{1}{\sqrt{2}}&-\frac{1}{\sqrt{2}}\\
0&-\frac{1}{\sqrt{2}}&\frac{1}{\sqrt{2}}
\end{pmatrix} \\
=&\begin{pmatrix}1&-1&0\\-1&0&1\end{pmatrix}\tag{21}
\end{align}

となり\(\mathsf{A}\)と一致する。

 

\(\mathsf{B}\)について

 まず特異値を求める。

 \(\mathsf{B}\)では\(\mathsf{B}^{\text{T}}\mathsf{B}\)が2次元の正方行列、\(\mathsf{B}\mathsf{B}^{\text{T}}\)が3次元の正方行列となる。

 よって\(\mathsf{B}^{\text{T}}\mathsf{B}\)の固有値を求める方が手っ取り早い。

\begin{align}
\mathsf{B}^{\text{T}}\mathsf{B}=\begin{pmatrix}4&0\\0&2\end{pmatrix}\tag{22}
\end{align}

より、固有値を\(\lambda\)とすると特性方程式より

\begin{gather}
\text{det}[\mathsf{B}\mathsf{B}^{\text{T}}-\lambda\mathsf{I}]=0 \\
\text{det}\left[\begin{pmatrix}4-\lambda&0 \\ 0&2-\lambda\end{pmatrix}\right]=0 \\
(4-\lambda)(2-\lambda)=0 \\
\therefore \lambda=2,4\tag{23}
\end{gather}

となるため、特異値は\(\sigma=\sqrt{2},2\)となる。

 

 続いて(\ref{aat})を利用して、特異値ごとに右特異ベクトルを求める。

 \(\mathsf{B}^{\text{T}}\mathsf{B}\)は実対称行列であるため、固有ベクトル(右特異ベクトル)は正規直交基底をなすように取ることができ、

\begin{align}
\begin{matrix}
\mathsf{B}^{\text{T}}\mathsf{B}\vec{v}^{(1)}=4\vec{v}^{(1)}&&
\mathsf{B}^{\text{T}}\mathsf{B}\vec{v}^{(2)}=2\vec{v}^{(2)}\\
\begin{pmatrix}4&0\\0&2\end{pmatrix}\begin{pmatrix}v_{1}^{(1)}\\v_{2}^{(1)}\end{pmatrix}=4\begin{pmatrix}v_{1}^{(1)}\\v_{2}^{(1)}\end{pmatrix}&&
\begin{pmatrix}4&0\\0&2\end{pmatrix}\begin{pmatrix}v_{1}^{(2)}\\v_{2}^{(2)}\end{pmatrix}=2\begin{pmatrix}v_{1}^{(2)}\\v_{2}^{(2)}\end{pmatrix}\\ \\
\Downarrow && \Downarrow\\ \\
v_{1}^{(1)}\in\mathbb{R},v_{2}^{(1)}=0 &&
v_{1}^{(2)}=0,v_{2}^{(2)}\in\mathbb{R} \\ \\
\end{matrix}\tag{24}
\end{align}

より、右特異ベクトルは

\begin{align}
\vec{v}^{(1)}=\begin{pmatrix}1\\0\end{pmatrix},\quad
\vec{v}^{(2)}=\begin{pmatrix}0\\1\end{pmatrix} \label{bhidari}\tag{25}
\end{align}

と求められる。

 

 続いて左特異ベクトルを求める。

 まず得られた特異値と右特異ベクトルを利用すれば、(\ref{tokui1})を変形して

\begin{align}
\vec{u}^{(1)}&=\frac{1}{2}\mathsf{A}\vec{v}^{(1)} &
\vec{u}^{(2)}&=\frac{1}{\sqrt{2}}\mathsf{A}\vec{v}^{(2)} \\
&=\frac{1}{2}\begin{pmatrix}-2&0\\0&1\\0&-1\end{pmatrix}\begin{pmatrix}1\\0\end{pmatrix} &
&=\frac{1}{\sqrt{2}}\begin{pmatrix}-2&0\\0&1\\0&-1\end{pmatrix}\begin{pmatrix}0\\1\end{pmatrix} \\
&=\begin{pmatrix}-1\\0\\0\end{pmatrix}&
&=\frac{1}{\sqrt{2}}\begin{pmatrix}0\\1\\-1\end{pmatrix}\tag{26}
\end{align}

と2つの左特異ベクトルが得られる。

 ここで、左特異ベクトルが\(\mathsf{B}\mathsf{B}^{\text{T}}\)の固有ベクトルであることを思い出すと、\(\mathsf{B}\mathsf{B}^{\text{T}}\)も実対称行列であるため、左特異ベクトルも正規交基底をなすように取ることができる。

 現に、\(\vec{u}^{(1)}\)と\(\vec{u}^{(1)}\)は互いに直交する単位ベクトルである。

 よってもう1つの左特異ベクトル\(\vec{u}^{(3)}\)は、左特異ベクトルも正規直交基底をなすことを利用して

\begin{gather}
\begin{cases}
\langle\vec{u}^{(1)},\vec{u}^{(3)}\rangle=0 \\
\langle\vec{u}^{(2)},\vec{u}^{(3)}\rangle=0 \\
|\vec{u}^{(3)}|=1
\end{cases} \\
\Downarrow \\
\begin{cases}
u_{1}^{(3)}=0 \\
u_{2}^{(3)}-u_{3}^{(3)}=0 \\
\sqrt{\left\{v_{1}^{(3)}\right\}^{2}+\left\{v_{2}^{(3)}\right\}^{2}+\left\{v_{3}^{(3)}\right\}^{2}}=1
\end{cases} \\
\Downarrow \\
\vec{u}^{(3)}=\frac{1}{\sqrt{2}}\begin{pmatrix}0\\1\\1\end{pmatrix}\tag{27}
\end{gather}

と求められる。

 

 以上より、

\begin{gather}
\Sigma=\begin{pmatrix}2&0\\0&\sqrt{2}\\0&0\end{pmatrix}\tag{28}\\
\mathsf{U}=(\vec{u}^{(1)},\vec{u}^{(2)},\vec{u}^{(3)})=\begin{pmatrix}
-1&0&0\\
0&\frac{1}{\sqrt{2}}&\frac{1}{\sqrt{2}}\\
0&-\frac{1}{\sqrt{2}}&\frac{1}{\sqrt{2}}
\end{pmatrix}\tag{29}\\
\mathsf{V}=(\vec{v}^{(1)},\vec{v}^{(2)})=\begin{pmatrix}
1&0\\
0&1\end{pmatrix}\tag{30}
\end{gather}

となるため、(\ref{tokuichilast})より

\begin{gather}
\mathsf{B}=\mathsf{U}\Sigma\mathsf{V}^{\text{T}} \\
\boxed{\begin{pmatrix}
-2&0\\
0&1\\
0&-1
\end{pmatrix}=\begin{pmatrix}
-1&0&0\\
0&\frac{1}{\sqrt{2}}&\frac{1}{\sqrt{2}}\\
0&-\frac{1}{\sqrt{2}}&\frac{1}{\sqrt{2}} \end{pmatrix}\begin{pmatrix}2&0\\0&\sqrt{2}\\0&0\end{pmatrix}
\begin{pmatrix}
1&0\\
0&1\end{pmatrix}}\label{btokuikai}\tag{31}
\end{gather}

と特異値分解できる。

 実際に(\ref{btokuikai})の右辺を計算すると、

\begin{align}
& \begin{pmatrix}
-1&0&0\\
0&\frac{1}{\sqrt{2}}&\frac{1}{\sqrt{2}}\\
0&-\frac{1}{\sqrt{2}}&\frac{1}{\sqrt{2}} \end{pmatrix}\begin{pmatrix}2&0\\0&\sqrt{2}\\0&0\end{pmatrix}
\begin{pmatrix}
1&0\\
0&1\end{pmatrix} \\
=&\begin{pmatrix}
-1&0&0\\
0&\frac{1}{\sqrt{2}}&\frac{1}{\sqrt{2}}\\
0&-\frac{1}{\sqrt{2}}&\frac{1}{\sqrt{2}} \end{pmatrix}\begin{pmatrix}2&0\\0&\sqrt{2}\\0&0\end{pmatrix} \\
=&\begin{pmatrix}
-2&0\\
0&1\\
0&-1
\end{pmatrix}\tag{32}
\end{align}

となり\(\mathsf{B}\)と一致する。

広告

終わりに

 とある理由で数学の勉強を始めたので、今後も新しく学ぶ領域に関しては記事にしていこうと思う。

 というか数学だけでなく、英語と国語も勉強しているという、さながら学生に戻ったかのような生活をしている。

 対価確保するようにしないと…

 

 END

広告

コメント

タイトルとURLをコピーしました