Numpy_3：使用数组进行面向数组编程

import numpy as np

使用数组进行面向数组编程

• 向量化：利用数组表达式来代替显式循环的方法
• 速度会快1～2个量级
• 广播机制是向量化的一种应用

np.meshgrid(arr_1, arr_2)

• 功能把输入的两个向量，匹配成可以一一组成的坐标点

points = np.arange(-5, 5, 0.01)

small_points_1 = np.array([1, 2, 3])
small_points_2 = np.array([7, 8])
np.meshgrid(small_points_1, small_points_2)

[array([[1, 2, 3],
        [1, 2, 3]]),
 array([[7, 7, 7],
        [8, 8, 8]])]

大型例子

x_s, y_s = np.meshgrid(points, points)
y_s

array([[-5.  , -5.  , -5.  , ..., -5.  , -5.  , -5.  ],
       [-4.99, -4.99, -4.99, ..., -4.99, -4.99, -4.99],
       [-4.98, -4.98, -4.98, ..., -4.98, -4.98, -4.98],
       ...,
       [ 4.97,  4.97,  4.97, ...,  4.97,  4.97,  4.97],
       [ 4.98,  4.98,  4.98, ...,  4.98,  4.98,  4.98],
       [ 4.99,  4.99,  4.99, ...,  4.99,  4.99,  4.99]])

z = np.sqrt(x_s**2 + y_s**2)
z

array([[7.07106781, 7.06400028, 7.05693985, ..., 7.04988652, 7.05693985,
       7.06400028],
       [7.06400028, 7.05692568, 7.04985815, ..., 7.04279774, 7.04985815,
        7.05692568],
       [7.05693985, 7.04985815, 7.04278354, ..., 7.03571603, 7.04278354,
        7.04985815],
       ...,
       [7.04988652, 7.04279774, 7.03571603, ..., 7.0286414 , 7.03571603,
        7.04279774],
       [7.05693985, 7.04985815, 7.04278354, ..., 7.03571603, 7.04278354,
        7.04985815],
       [7.06400028, 7.05692568, 7.04985815, ..., 7.04279774, 7.04985815,
        7.05692568]])

用matplotlib来尝试一下

import matplotlib.pyplot as plt

plt.imshow(z, cmap=plt.cm.gray); plt.colorbar()
plt.title("Image plot of $\sqrt{x^2 + y^2}$ for a grid of values")
plt.draw()

plt.close('all')

将条件逻辑作为数组操作

• np.where()是 x if CONDITION else y 的向量化表示

x_arr = np.array([1.1, 1.2, 1.3, 1.4, 1.5])
y_arr = np.array([2.1, 2.2, 2.3, 2.4, 2.5])
cond = np.array([True, False, True, True, False])

• 需求为：当 cond 为 True 时，取 x_arr 中元素，反之 y_arr

result = np.where(cond, x_arr, y_arr)
result

array([1.1, 2.2, 1.3, 1.4, 2.5])

• 随机生成4*4矩阵，将所有的正值替换为+2，负值替换为-2

matrix_4_4 = np.random.randn(4, 4)
matrix_4_4

array([[-0.31057548, -0.33117123, -0.32092367, -0.83744854],
       [ 0.65508715, -1.09744727, -0.28864518,  1.87957568],
       [ 0.13691023,  1.27582489, -0.73059565, -1.25889417],
       [ 0.94719682,  1.22910758, -1.44860514,  0.63306565]])

cond = matrix_4_4 > 0

result = np.where(cond, 2, -2)
result

array([[-2, -2, -2, -2],
       [ 2, -2, -2,  2],
       [ 2,  2, -2, -2],
       [ 2,  2, -2,  2]])

数学和统计方法

• 聚合函数，可以np.sum(arr)也可以使用arr.sum()

arr = np.random.randn(5, 4)
arr

array([[-6.15103319e-01,  1.94935393e-01,  1.07466373e+00,
         5.60804143e-01],
       [-1.29562464e+00, -9.50808332e-01, -6.91278641e-01,
        -1.90757357e-03],
       [ 2.14085181e+00, -5.65517145e-03,  2.99332159e-01,
        -4.74249005e-01],
       [-1.46233648e+00,  7.16732377e-01, -1.02124467e-01,
         3.93163196e-02],
       [-9.34729945e-01,  3.65241170e-01, -5.98246162e-02,
        -1.46758290e-01]])

arr.mean()

-0.06742616905151418

np.mean(arr)

-0.06742616905151418

arr.sum()

-1.3485233810302835

• 可以选轴的，axis

test = np.array([[1, 1, 1, 1],
          [2, 2, 2, 2],
          [3, 3, 3, 3]])

print(test.mean(axis = 0))
print(test.mean(axis = 1))

[2. 2. 2. 2.]
[1. 2. 3.]

• np.cumsum():从0开始元素累积和
• np.cumprod():从1开始元素累积积

arr = np.array([[1, 1, 2], 
                [3, 4, 5], 
                [6, 7, 8]])
arr

array([[1, 1, 2],
       [3, 4, 5],
       [6, 7, 8]])

arr.cumsum(axis=0)

array([[ 1,  1,  2],
       [ 4,  5,  7],
       [10, 12, 15]])

arr.cumprod(axis=1)

array([[  1,   1,   2],
       [  3,  12,  60],
       [  6,  42, 336]])

from PIL import Image

Image.open('./numpy_statistic.jpeg')

BOOL数组的方法

• bool值会被强制定为1 or 0
• 可以用来计算非0个数

arr = np.random.randn(100)

(arr>0).sum()

49

• np.any() 和 np.all()
• any 是至少有一个True嘛？
• all 是全是True嘛？

bools = np.array([False, False, True, False])

bools.any()

True

bools.all()

False

排序

• np.sort() 返回的是排好序数组拷贝，不会原数组改变顺序
• 根据某一列进行排序，会出现在pandas中

arr = np.random.randn(5, 3)
arr

array([[-0.71284458, -0.59405487,  1.20906567],
       [-1.25745017,  0.90948517, -0.24589889],
       [-0.18246322, -1.85378669, -1.14372438],
       [ 0.19665606, -2.42751683,  1.75191219],
       [ 0.07381909, -0.27235433, -0.74304361]])

np.sort(arr, 0)

array([[-1.25745017, -2.42751683, -1.14372438],
       [-0.71284458, -1.85378669, -0.74304361],
       [-0.18246322, -0.59405487, -0.24589889],
       [ 0.07381909, -0.27235433,  1.20906567],
       [ 0.19665606,  0.90948517,  1.75191219]])

唯一值与其他集合逻辑

• np.unique() 返回数组中唯一值的排序好的数组

names = np.array(['Bob', 'Joe', 'Will', 'Bob', 'Will', 'Joe', 'Joe'])
names

array(['Bob', 'Joe', 'Will', 'Bob', 'Will', 'Joe', 'Joe'], dtype='<U4')

np.unique(names)

array(['Bob', 'Joe', 'Will'], dtype='<U4')

• np.in1d([], []) 可以检查一个数组是否在另一个数组中，返回一个bool数组

values = np.array([6, 0, 0, 3, 2, 5, 6])

np.in1d(values, [2, 3, 6])

array([ True, False, False,  True,  True, False,  True])

Image.open('./List_set.jpeg')