0
0

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?

More than 1 year has passed since last update.

『スタンフォード ベクトル・行列からはじめる最適化数学』9章章末問題の解答

Last updated at Posted at 2023-01-09

スタンフォード ベクトル・行列からはじめる最適化数学

の演習問題を解いているが解答が本書内にもネット上にもない模様。自分の解答を晒して間違いにツッコミをいただき理解を深めようという試み。

- 問9.1

\begin{align}
\begin{bmatrix}
0.9 & 0    & 0.05 \\
0.1 & 0.95 & 0 \\
0   & 0.05 & 0.92 \\
\end{bmatrix}
\begin{bmatrix}
x_{t_1} \\
x_{t_2} \\
x_{t_3} \\
\end{bmatrix}
&=
\begin{bmatrix}
x_{(t+1)_1} \\
x_{(t+1)_2} \\
x_{(t+1)_3} \\
\end{bmatrix} \\
x_{(t+1)_1}&=0.9x_{t_1}\quad\quad\quad+0.05x_{t_3} \\
x_{(t+1)_2}&=0.1x_{t_1}+0.95x_{t_2} \\
x_{(t+1)_3}&=\quad\quad\quad0.05x_{t_2}+0.95x_{t_3} \\
\end{align}
  • 問9.2
    • (a) セクター3のt期がセクター2のt+1期の産出量に及ぼす影響の度合い。
    • (b)
import numpy as np
import matplotlib.pyplot as plt

a = np.array([0.6, 0.9, 1.3, 0.5])
B = np.array([[0.1, 0.06, 0.05, 0.70], [0.48, 0.44, 0.1, 0.04], [0.0, 0.55, 0.52, 0.04], [0.04, 0.01, 0.42, 0.51]])

sum_a = []
sum_a.append(sum(a))
for i in range(1, 20):
  a = B @ a
  sum_a.append(sum(a))

plt.figure(figsize=(12, 6))
plt.plot(list(range(1, 21)), sum_a)

plt.xlabel('time')
plt.ylabel('sum_a')
plt.show()

image.png

  • 問9.3
\begin{align}
(I-A)z&=c \\
z&=(I-A)^{-1}c \\
また、\\
Fz&=g \\
z&=F^{-1}g \\
より、\\
F&=I-A\\
g&=c
\end{align}
  • 問9.4
\begin{align}
x_{t+1}&=A_1x_t+A_2x_{t-1} \\
z_t&=
\begin{bmatrix}
x_t \\
x_{t-1} \\
\end{bmatrix}\\
z_{t+1}&=
\begin{bmatrix}
x_{t+1} \\
x_t \\
\end{bmatrix}\\
&=
\begin{bmatrix}
A_1 & A_2 \\
1   & 0 \\
\end{bmatrix}
\begin{bmatrix}
x_t \\
x_{t-1} \\
\end{bmatrix}\\
よって、B&=
\begin{bmatrix}
A_1 & 0 \\
0   & A_2 \\
\end{bmatrix}となるBを考えれば\\
z_{t+1}&=Bz_tとなる。
\end{align}
  • 問9.5
\begin{align}
問9.4より、\\
x_{t+1}&=A_1x_t+A_2x_{t-1}\\
z_t&=(x_t, x_{t-1}) \\
\iff z_{t+1}&=Bz_t\\
y_{t+1}&=y_t+y_{t-1} \\
x_t&=(y_t, y_{t-1}) \\
x_{t+1}&=Bx_t \\
B&=\begin{bmatrix}
1 & 1 \\
1 & 0 \\
\end{bmatrix}
\end{align}
import numpy as np

x = np.array([0, 1])
B = np.array([[1, 1], [1, 0]])
print(x[0])
for i in range(20):
  x = B @ x
  print(x[0])
0
1
1
2
3
5
8
13
21
34
55
89
144
233
377
610
987
1597
2584
4181
6765
  • 修正フィボナッチ数列
\begin{bmatrix}
z_t \\
z_{t-1} \\
\end{bmatrix}
=
\begin{bmatrix}
1 & -1 \\
1 & 0 \\
\end{bmatrix}
\begin{bmatrix}
z_{t-1} \\
z_{t-2} \\
\end{bmatrix}
import numpy as np

z = np.array([0, 1])
B = np.array([[1, -1], [1, 0]])
print(z[0])
for i in range(20):
  z = B @ z
  print(z[0])
0
-1
-1
0
1
1
0
-1
-1
0
1
1
0
-1
-1
0
1
1
0
-1
-1
  • 問9.6
\begin{align}
x_t&=\frac{u_1+u_2+\cdots+u_{t-1}}{t-1} \\
x_{t+1}&=\frac{u_1+u_2+\cdots+u_{t-1}+u_t}{t} \\
&=\frac{t-1}{t}\left(\frac{u_1+u_2+\cdots+u_{t-1}}{t-1}+\frac{u_t}{t-1}\right) \\
&=\frac{t-1}{t}\left(x_t+\frac{u_t}*{t-1}\right) \\
&=\frac{t-1}{t}x_t+\frac{u_t}{t}

\end{align}
  • 問9.7
X_{t+1}=AX_t+Bu_t \\
t 1単位につき、計算量は\\
(n+(n-1))\times n \\
(m+(m-1))\times m \\
(2n-1)n+(2m-1)m \\
が10^5回あるから\\
{(2\times15-1)\times 15+(2\times 5 - 0.5)}\times10^5 \\
4.85\times10^7 \\
\frac{4.85\times10^7}{10^9}=4.85\times10^{-2}s

0
0
0

Register as a new user and use Qiita more conveniently

  1. You get articles that match your needs
  2. You can efficiently read back useful information
  3. You can use dark theme
What you can do with signing up
0
0

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?