Skip to content

京都大学 情報学研究科 知能情報学専攻 2023年8月実施 情報学基礎 F1-1

Author

Isidore

Description

以下の設問において \(i\) は虚数単位を, \(\mathbb{C}\) は複素数全体の集合を表す。 また、行列 \(\boldsymbol{A}\) に対して、\(\boldsymbol{A}^H\)\(\boldsymbol{A}\) の共役転置(エルミート転置)を、\(\boldsymbol{A}^{-1}\)\(\boldsymbol{A}\) の逆行列をそれぞれ表す。

設問1

行列 \(\boldsymbol{D} \in \mathbb{C}^{4 \times 4}\) を次で定義する。

\[ \boldsymbol{D} = \frac{1}{2} \begin{pmatrix} 1 & 1 & 1 & 1 \\ 1 & -i & -1 & i \\ 1 & -1 & 1 & -1 \\ 1 & i & -1 & -i \end{pmatrix} \]

このとき、以下の問いに答えよ。

(1) \(\boldsymbol{D}\) がユニタリ行列であることを示せ。

(2) 行列 \(\boldsymbol{G} \in \mathbb{C}^{4 \times 4}\)

\[ G = \begin{pmatrix} 1 & 0 & 0 & 0 \\ 0 & i & 0 & 0 \\ 0 & 0 & 1 & 0 \\ 0 & 0 & 0 & i \end{pmatrix} \]

で定義するとき、\(\boldsymbol{D}^H \boldsymbol{G} \boldsymbol{D}\) の逆行列を導出せよ。

設問2

\(n\) 次元ベクトル空間 \(V\) が2つの部分空間 \(W_1, W_2\) の直和であるとする。ベクトル \(x \in V\)

\[ x = x_1 + x_2, \quad x_1 \in W_1, x_2 \in W_2 \]

と分解されるとき、\(x\)\(x_1\) に写す一次変換を考える。 この一次変換を表す \(V\) のある基底に関する行列を \(\boldsymbol{B}\) とする。このとき、以下の問いに答えよ。

(1) \(x\)\(x_1\) に写す一次変換を表す上述の基底に関する行列を導出せよ。

(2) \(\boldsymbol{B}^2 = \boldsymbol{B}\) が成り立つことを示せ。

(3) 適当な \(n\) 次正則行列 \(\boldsymbol{P}\) をとれば

\[ \boldsymbol{P}^{-1} \boldsymbol{B} \boldsymbol{P} = \begin{pmatrix} 1 & & & & & \\ & \ddots & & & & \\ & & 1 & & & \\ & & & 0 & & \\ & & & & \ddots & \\ & & & & & 0 \end{pmatrix} \]

となることを示せ。

Kai

設問1

(1)

Definition of Unitary Matrix: \(D^{H}D=E\), in which \(E\) is identity matrix and \(D^H\) stands for Hermitian Matrix. Calculation omitted.

(2)

\[ \begin{aligned} (D^{H}GD)^{-1} &= (D^{-1}G^{-1}(D^H)^{-1}) = D^{H}G^{-1}D \\ & =\begin{bmatrix} 2-2i & 0 & 2+2i & 0 \\ 0 & 2-2i & 0 & 2+2i \\ 2+2i & 0 & 2-2i & 0 \\ 0 & 2+2i & 0 & 2-2i \end{bmatrix} \end{aligned} \]

設問2

(1)

According to the question, the linear transformation \(B\) maps \(x\) to \(x_1\), so

\[ \begin{align} Bx=x_1 \tag{*} \end{align} \]

Insert \(x_1 = x-x_2\), we get

\[ Bx=x - x_2 \Rightarrow (E - B)x = x_2 \]

So \(E-B\) is the answer.

(2)

Insert \(x=x_1 + x_2\) into equation (*),

\[ Bx_1 + Bx_2 = x_1 \]

Given \(V = W_1 \oplus W_2\), by the properties of Projection Matrix we have

\[ Bx_1 = x_1, Bx_2 = 0 \]

Then we have

\[ B^2x_1 = Bx_1 = x_1, B^2x_2 = Bx_2 = 0 \]

Therefore, \(\forall x \in V, \; B^2x= x_1 = Bx\), which implies that \(B^2 = B\).

(3)

Consider the eigenvalues \(\lambda\) of \(B\), which satisfies

\[ Bx = \lambda x \]

By (2) we have \(B^2 = B\), hence every \(\lambda\) satisfies

\[ \lambda^2 = \lambda \]

Hence \(\lambda = 1,0\).

Assume that the algebraic multiplicity of \(\lambda = 1\) is \(k\), then the algebraic multiplicity of \(\lambda = 0\) is \((n-k)\).

Therefore, we can derive

\[ P=\begin{bmatrix}x_1 & x_2 & ...... & x_n \end{bmatrix} \]

in which \(x_1...x_k\) is the eigenvectors corresponding to \(\lambda = 1\) and the others is the eigenvectors corresponding to \(\lambda = 0\).

Consider all the \(Bx_i=\lambda x_i\), we can get

\[ BP = P\begin{bmatrix} E & 0 \\ 0 & 0 \end{bmatrix}\]

We give a brief check of whether \(P\) is non-singular or not.

When \(\lambda = 1\),

\[ Bx_i=x_i \Rightarrow (E-B)x_i = 0 \]

So its eigenspace is \(N(E-B)\). Similarly, \(\lambda = 0\)'s eigenspace is \(N(B)\).

Obviously, \(N(B) = W_2\) and then \(N(E-B) = W_1\). Given \(V = W_1 \oplus W_2\), the space spanned by column vectors of \(P\) (which are eigenvectors of \(B\)) is exact \(V\). Therefore, \(P\) is non-singular. Q.E.D.