Numpy

import numpy as np
Numpy 基本運算 (1/2)
l NumPy 是 Python 語⾔的⼀個第三⽅函式庫，⽀援多維度陣 import numpy as np import numpy as np
列與矩陣運算，Numpy 具備以下特性： # creat matrix matrix1 = np.array([[1, 2],
matrix1 = np.array([ [3, 4],])
Ø ⽀援多維度陣列類別資料 [1, 2, 3], matrix2 = np.array([[5, 6],
[4, 5, 6], [7, 8],])
Ø 廣播運算的特性，使矩陣運算容易理解執⾏ [7, 8, 9],])
foo = [['a', 'b', 'c'], a = matrix1 + matrix2
Ø 底層由 C/C++ 和 Fortran 實作，執⾏效率佳 ['d', 'e', 'f'], b = matrix1 - matrix2
['g', 'h', 'i'],]
Ø ⼤量科學計算的實作函數，如線性代數、三⾓函數等 matrix2 = np.array(foo) # 對應位置的乘積 c1, c2
boo = matrix1[1][1] c1 = matrix1 * matrix2
l NumPy的矩陣是 ndarray 類別的物件 (N維陣列, n- zoo = matrix2[1,2] c2 = np.multiply(matrix1, matrix2)
dimensional array), 可以⽤多層list來建立, 只能儲存相同資 # create 2X2 (two-dimension) matrix
l22 = [[1, 2], d = matrix1 / matrix2
料型別的物件 [3, 4],]
m22 = np.array(l22) # 矩陣乘法 e1, e2
l ndarray 運算⼦: +, -, *, /, @ egg = m22[1,1]
# create 3X3 (two-dimension) matrix
e1 = matrix1 @ matrix2
e2 = np.matmul(matrix1, matrix2)
l ndarray 內建函式: l33 = [[1, 2, 3],
[4, 5, 6], # 內積: 2D相當於@, in 1D 計算為⼀個純量
[7, 8, 9],] f1 = np.dot(matrix1, matrix2)
• array() m33 = np.array(l33) f2 = np.dot(np.array([1, 3]), np.array([2, 4]))
spam = m33[1,1]
• dot() 內積: in 2D相當於@, in 1D 計算為⼀個純量 # create 2X3X3 (three-dimension) matrix
l233 = [[[1, 2, 3],
• multiply() 對應位置的乘積 (*) [4, 5, 6],
• matmul() 矩陣乘法 (@): A@B is equal to np.matmul(A, B) [7, 8, 9]],
[[10, 11, 12],
[13, 14, 15],
[16, 17, 18]],]
m233 = np.array(l233)
hen = m233[1,1,1]
Numpy 基本運算 (2/2)
import numpy as np import numpy as np
A = np.array([[1, 2, 3], A = np.array([[1, 2, 3],
l ndarray 內建函式: [4, 5, 6],]) [4, 5, 6],
B = np.array([[7, 8, 9], [7, 8, 9],])
ü vstack(), hstack()
[10, 11, 12],])
ü reshape() a = A.shape
C = np.vstack([A, B]) b = A.diagonal()
ü arange()
D = np.hstack([A, B]) c = A.flatten()
l ndarray 型別⽅法: E = A.reshape(6, 1) d = A.transpose()
F = A.reshape(1, 6) e = A.min()
ü .shape G = A.reshape(3, 2) f = A.max()
g = A.mean()
ü .diagonal(), .flatten(), .transpose()
# create matrix by numpy h = A.sum()
ü .min(), .max(), .mean(), .sum() nums = np.arange(1, 5) foo = A[0][1]
H = nums.reshape(2, 2) boo = A[0, 1]
l ndarray indexing: matrix[0][1], matrix[0, 1] I = np.arange(1, 13).reshape(3, 2, 2)
J = np.arange(1, 24, 2).reshape(3, 2, 2)
#Lab

NumPy 運算
1. use arrange() and reshape() 產⽣右⽅的matrix
2. ⽤ np.arange() 和 np.reshape(), 建立⼀個3x3 Numpy 陣列A, 內容是數字 3 到 11
• 求陣列 A 的所有元素的最⼩值，最⼤值，平均值，總和
• 將 A 的每個元素進⾏平⽅計算，並將結果存到 B
3. 設 a = (78, 22, 65, 87, 12, 98, 63, 79) 且 x = 57, 試著在a 中找出離 x 最近的數，請寫兩個版本，⼀個⽤for, 另
⼀個⽤Numpy廣播運算

#1 #2 #3A #3B
import numpy as np import numpy as np import numpy as np import numpy as np
a = np.arange(1,10) A = np.arange(3,12).reshape(3, 3) import math a = (78, 22, 65, 87, 12, 98, 63, 79)
b = np.arange(11,20) a = A.shape a = (78, 22, 65, 87, 12, 98, 63, 79) x = 57
c = np.arange(21,30) b = A.min() diff = () A = np.array(a)
d = np.hstack([a, b]) c = A.max() x = 57 B = abs(A - 57)
e = np.hstack([d, c]) d = A.mean() min_index = 0 min_distance = min(B)
matrix_1 = e.reshape(3, 3, 3) e = A.sum() min_distance = math.inf pos = np.where(B == min_distance)[0][0]
B=A*A for i in range(0, len(a)): num = a[pos]
z = np.arange(1, 30) if min_distance > abs(x-a[i]): print(f'a[{pos}]:{a[pos]} have min value
y = np.delete(z, [9, 19]) min_index = i {min_distance}')
matrix_2 = y.reshape(3, 3, 3) min_distance = abs(x-a[i])

print(f'a[{min_index}]:{a[min_index]}
have min value {min_distance}')