Numpy学习笔记

1、数组的定义和应用

2、数组元素的索引选取

3、数组的计算

4、线性代数的运行计算

1 Arrays

array的维数就是array的秩（rank)

import numpy as np
a = np.array([1, 3, 4, 6])
# 此时是列向量

print(a)

[1 3 4 6]

print(a.size)
print(a.shape)

4
(4,)

三维数组

b = np.array([[[1,2,3],[1,2,3]]])
print(b)
print('---------')
print(b.shape)

[[[1 2 3]
  [1 2 3]]]
---------
(1, 2, 3)

2、创建Array

c = np.zeros([2,3])

print(c)

[[0. 0. 0.]
 [0. 0. 0.]]

zeros_like : 仿照一个数组格式创造一个全0数组

d = np.zeros_like(b)

print(d)

[[[0 0 0]
  [0 0 0]]]

eye(2) : 创建 i

e = np.eye(3)
print(e)

[[1. 0. 0.]
 [0. 1. 0.]
 [0. 0. 1.]]

3、Array的常用属性和方法

random.rand(3,4)：返回 3 行 4 列的 0-1 之间随机值

f = np.random.rand(3,4)
f

array([[0.29918133, 0.99695443, 0.70050163, 0.16706489],
       [0.44481688, 0.55517105, 0.9650693 , 0.93577429],
       [0.70169262, 0.95443642, 0.52026445, 0.0306768 ]])

SUM：求和的 3 种方法

np.sum()和print(np.sum())区别在输出是否有array()

np.sum(f,axis=0)

array([1.44569082, 2.5065619 , 2.18583537, 1.13351597])

axis=0 /1 分别代表对列和对行操作

print(np.sum(f))
print(np.sum(f,axis=0))
print(np.sum(f,axis=1))

7.271604071001589
[1.44569082 2.5065619  2.18583537 1.13351597]
[2.16370227 2.90083152 2.20707029]

sort

# 默认是按行穿透排序,即axis=1
np.sort(f)

array([[0.16706489, 0.29918133, 0.70050163, 0.99695443],
       [0.44481688, 0.55517105, 0.93577429, 0.9650693 ],
       [0.0306768 , 0.52026445, 0.70169262, 0.95443642]])

np.sort(f,axis=0)

array([[0.29918133, 0.55517105, 0.52026445, 0.0306768 ],
       [0.44481688, 0.95443642, 0.70050163, 0.16706489],
       [0.70169262, 0.99695443, 0.9650693 , 0.93577429]])

# 返回的是元素大小的索引
f.argsort()

array([[3, 0, 2, 1],
       [0, 1, 3, 2],
       [3, 2, 0, 1]])

# 返回对应行/列的最大值索引
f.argmax(axis=1)

array([1, 2, 1])

# 返回大于0.5的索引，数组1是行数，数组2是列数，对应着看
np.where(f>0.5)

(array([0, 0, 1, 1, 1, 2, 2, 2]), array([1, 2, 1, 2, 3, 0, 1, 2]))

4、随机数

rand(d0, d1, d2 …… dn) 输出0-1之间的随机数，每个参数是一维

randn(d0, d1, d2 …… dn) 输出正态分布的0-1随机数，每个参数是一维

randint(low, high, size,dtype) 输入下上界，维数，*

choice(a, size) 输入参数可以是上界/数组（限定从中筛选）

生成0-1之间的随机数

np.random.rand(10)

array([0.71028704, 0.43191226, 0.50255482, 0.21935992, 0.39390923,
       0.57976832, 0.9898773 , 0.95256839, 0.33815693, 0.89487692])

np.random.rand(3,4)

array([[0.03521269, 0.68399925, 0.00300619, 0.13844359],
       [0.6677425 , 0.54130992, 0.34670055, 0.6620398 ],
       [0.87538985, 0.7452303 , 0.5317525 , 0.95381903]])

生成正态分布的随机数

np.random.randn(10)

array([ 1.31280276, -0.10793412, -0.73151144, -2.25348606,  1.04692662,
        0.53299785, -0.28173337,  1.37021352,  1.33366902,  1.62139734])

生成范围内随机数组

np.random.randint(4,7,size=3)

array([5, 4, 5])

np.random.randint(9,15,size=(3,4))

array([[10, 11, 11,  9],
       [ 9, 14, 14, 13],
       [14, 13, 10,  9]])

random.random()和random.rand()没有明显区别

np.random.random(10)

array([0.69736939, 0.35938908, 0.05222814, 0.36664614, 0.45729116,
       0.92235331, 0.00982576, 0.48825766, 0.52762227, 0.18377659])

random.choice()和random.randint()的区别

randint()的前两个参数是下上界，choice的第一个参数可以是数组和上界

np.random.choice([1,2,3],(3,4))

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

np.random.rand(2,4)

array([[0.15795165, 0.12727261, 0.5790727 , 0.14881722],
       [0.95894915, 0.60085816, 0.28078072, 0.86660551]])

5、数组的索引

g = np.random.randint(1,10,(6,7))
g

array([[2, 8, 8, 2, 4, 3, 3],
       [8, 9, 7, 9, 1, 5, 1],
       [3, 4, 8, 6, 9, 5, 9],
       [6, 2, 4, 7, 2, 1, 2],
       [1, 4, 8, 1, 5, 4, 7],
       [6, 1, 9, 6, 2, 5, 9]])

切片

g[0:2,0:3]

array([[2, 8, 8],
       [8, 9, 7]])

布尔型索引

g>4

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

g[g>7]

array([8, 8, 8, 9, 9, 8, 9, 9, 8, 9, 9])

g[2::2,::2]

array([[3, 8, 9, 9],
       [1, 8, 5, 7]])

6、数组的数学

h = np.random.random([3,4])
i = np.random.random([3,4])

h

array([[0.31145957, 0.6485324 , 0.62999148, 0.34106203],
       [0.08457065, 0.20985707, 0.20120685, 0.17891792],
       [0.50814369, 0.12085831, 0.87225719, 0.35348592]])

h+2

array([[2.31145957, 2.6485324 , 2.62999148, 2.34106203],
       [2.08457065, 2.20985707, 2.20120685, 2.17891792],
       [2.50814369, 2.12085831, 2.87225719, 2.35348592]])

h+i

array([[1.14708914, 0.89204308, 1.19424484, 0.6964331 ],
       [0.74506304, 0.38368983, 0.63390267, 0.29182269],
       [1.40465142, 1.10406526, 1.40642566, 1.02236906]])

# 矩阵的内积
np.dot(a,b.T) #或者a.dot(b.T)

array([[1.64305069, 1.03716829, 1.50919486],
       [1.42773865, 1.13712743, 1.37372482],
       [1.06103007, 0.6484528 , 0.75675077]])