numpy get started
# python: 一切皆对象
# linux: 一切皆文件.
# 数据分析: 一切皆矩阵
导入numpy库,并查看numpy版本
import numpy as np
np.__version__
'1.17.0'
如果没有安装相关包:
!pip3 install matplotlib -i https://pypi.douban.com/simple
!pip3 install pillow -i https://pypi.douban.com/simple
import matplotlib.pyplot as plt
# 读取图片
cat = plt.imread("cat.jpg")
numpy 里面的矩阵不是真正的矩阵, 只是numpy用来表示矩阵的一种数据结构, 叫做ndarray
cat是一个彩色图片.
jpg的结尾的图片,里面RGB颜色值都是0到255的整数.
png结尾的图片,里面RBG颜色值都是0到1的浮点数.
cat # 三维的
array([[[231, 186, 131],
[232, 187, 132],
[233, 188, 133],
...,
[100, 54, 54],
[ 92, 48, 47],
[ 85, 43, 44]],
[[232, 187, 132],
[232, 187, 132],
[233, 188, 133],
...,
[100, 54, 54],
[ 92, 48, 47],
[ 84, 42, 43]],
[[232, 187, 132],
[233, 188, 133],
[233, 188, 133],
...,
[ 99, 53, 53],
[ 91, 47, 46],
[ 83, 41, 42]],
...,
[[199, 119, 82],
[199, 119, 82],
[200, 120, 83],
...,
[189, 99, 65],
[187, 97, 63],
[187, 97, 63]],
[[199, 119, 82],
[199, 119, 82],
[199, 119, 82],
...,
[188, 98, 64],
[186, 96, 62],
[188, 95, 62]],
[[199, 119, 82],
[199, 119, 82],
[199, 119, 82],
...,
[188, 98, 64],
[188, 95, 62],
[188, 95, 62]]], dtype=uint8)
print(cat)
[[[231 186 131]
[232 187 132]
[233 188 133]
...
[100 54 54]
[ 92 48 47]
[ 85 43 44]]
[[232 187 132]
[232 187 132]
[233 188 133]
...
[100 54 54]
[ 92 48 47]
[ 84 42 43]]
[[232 187 132]
[233 188 133]
[233 188 133]
...
[ 99 53 53]
[ 91 47 46]
[ 83 41 42]]
...
[[199 119 82]
[199 119 82]
[200 120 83]
...
[189 99 65]
[187 97 63]
[187 97 63]]
[[199 119 82]
[199 119 82]
[199 119 82]
...
[188 98 64]
[186 96 62]
[188 95 62]]
[[199 119 82]
[199 119 82]
[199 119 82]
...
[188 98 64]
[188 95 62]
[188 95 62]]]
type(cat)
numpy.ndarray
plt.imshow(cat)
<matplotlib.image.AxesImage at 0x7fccb0da0908>
cat.shape # 三维的
(456, 730, 3)
#请问电影是什么,nd.array 四维
#(x,456,760,3)
一、创建ndarray
1. 使用np.array()由python list创建
参数为列表:
[1, 4, 2, 5, 3]
注意:
- numpy默认ndarray的所有元素的类型是相同的
- 如果传进来的列表中包含不同的类型,则统一为同一类型,优先级:str>float>int
l = [1, 4, 2, 5, 3]
n = np.array(l)
n
array([1, 4, 2, 5, 3])
l = [1, 4, 2, '5', 3.0]
n = np.array(l)
n
array(['1', '4', '2', '5', '3.0'], dtype='<U21')
n[0] = 2
n
array(['2', '4', '2', '5', '3.0'], dtype='<U21')
l
[1, 4, 2, '5', 3.0]
l[0] = 10
l
[10, 4, 2, '5', 3.0]
n
array(['2', '4', '2', '5', '3.0'], dtype='<U21')
n.shape
(5,)
n2 = np.array([[3, 4, 7, 1], [3, 0, 1, 6], [4, 0, 1, 4]])
n2
array([[3, 4, 7, 1],
[3, 0, 1, 6],
[4, 0, 1, 4]])
n2.shape
(3, 4)
lis = [[[1,2,3,4], [3,4,5,6]], [[1,2,3,4], [3,4,5,6]]]
n3 = np.array(lis)
n3
array([[[1, 2, 3, 4],
[3, 4, 5, 6]],
[[1, 2, 3, 4],
[3, 4, 5, 6]]])
n3.shape
(2, 2, 4)
2. 使用np的routines函数创建
包含以下常见创建方法:
1) np.ones(shape, dtype=None, order='C')
# 用来产生全是1的ndarray
# shape是用来指定ndarray的形状
# dtype是数字的类型
np.ones(shape=(4, 8), dtype=np.int8)
array([[1, 1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1, 1]], dtype=int8)
n2 = np.ones((2, 3, 4), dtype=int)
n2
array([[[1, 1, 1, 1],
[1, 1, 1, 1],
[1, 1, 1, 1]],
[[1, 1, 1, 1],
[1, 1, 1, 1],
[1, 1, 1, 1]]])
2) np.zeros(shape, dtype=float, order='C')
# 全是0 的ndarray
np.zeros(shape=(4, 3), dtype=np.float32)
array([[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.]], dtype=float32)
n2 = np.zeros((5, 5))
n2
array([[0., 0., 0., 0., 0.],
[0., 0., 0., 0., 0.],
[0., 0., 0., 0., 0.],
[0., 0., 0., 0., 0.],
[0., 0., 0., 0., 0.]])
3) np.full(shape, fill_value, dtype=None, order='C')
# 按照指定数字产生ndarray
np.full((3, 5), fill_value=8.)
array([[8., 8., 8., 8., 8.],
[8., 8., 8., 8., 8.],
[8., 8., 8., 8., 8.]])
4) np.eye(N, M=None, k=0, dtype=float)
对角线为1其他的位置为0
# 生成对角线全是1,其他位置全是0的2维ndarray
# k=0 表示主对角线的索引.
# 像这种主对角线全是1, 其他位置全是0的矩阵,我们叫做单位矩阵.
# 任何矩阵乘于他的单位矩阵,等于他本身.
# 如果一个矩阵可以通过矩阵的行列式变化为单位矩阵,那么说明这个矩阵有唯一的解.
np.eye(6)
array([[1., 0., 0., 0., 0., 0.],
[0., 1., 0., 0., 0., 0.],
[0., 0., 1., 0., 0., 0.],
[0., 0., 0., 1., 0., 0.],
[0., 0., 0., 0., 1., 0.],
[0., 0., 0., 0., 0., 1.]])
np.eye(6, k=1)
array([[0., 1., 0., 0., 0., 0.],
[0., 0., 1., 0., 0., 0.],
[0., 0., 0., 1., 0., 0.],
[0., 0., 0., 0., 1., 0.],
[0., 0., 0., 0., 0., 1.],
[0., 0., 0., 0., 0., 0.]])
np.eye(3, M=5, dtype=int)
array([[1, 0, 0, 0, 0],
[0, 1, 0, 0, 0],
[0, 0, 1, 0, 0]])
5) np.linspace(start, stop, num=50, endpoint=True, retstep=False, dtype=None)
# endpoint是否包含stop的值
# retstep是否返回步长.
n = np.linspace(0, 100, num=50, dtype=int, retstep=True, endpoint=False)
n
(array([ 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32,
34, 36, 38, 40, 42, 44, 46, 48, 50, 52, 54, 56, 58, 60, 62, 64, 66,
68, 70, 72, 74, 76, 78, 80, 82, 84, 86, 88, 90, 92, 94, 96, 98]), 2.0)
np.linspace(0, 150, num=30, dtype=np.int8)
array([ 0, 5, 10, 15, 20, 25, 31, 36, 41, 46, 51,
56, 62, 67, 72, 77, 82, 87, 93, 98, 103, 108,
113, 118, 124, -127, -122, -117, -112, -106], dtype=int8)
6) np.arange([start, ]stop, [step, ]dtype=None)
range(0, 10)
range(0, 10)
# 就把np.arange当成numpy版本的range
# 使用np.arange的时候,尽量用整形参数.
np.arange(50)
array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16,
17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33,
34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49])
np.arange(0.1, 11.1, step=0.35)
array([ 0.1 , 0.45, 0.8 , 1.15, 1.5 , 1.85, 2.2 , 2.55, 2.9 ,
3.25, 3.6 , 3.95, 4.3 , 4.65, 5. , 5.35, 5.7 , 6.05,
6.4 , 6.75, 7.1 , 7.45, 7.8 , 8.15, 8.5 , 8.85, 9.2 ,
9.55, 9.9 , 10.25, 10.6 , 10.95])
7) np.random.randint(low, high=None, size=None, dtype='l')
np.random.randint(10)
9
# 左闭右开区间
n = np.random.randint(0, 100, size=(3, 4, 5))
n
array([[[95, 26, 25, 0, 6],
[63, 91, 8, 6, 0],
[28, 62, 80, 20, 69],
[60, 0, 75, 3, 99]],
[[52, 36, 55, 72, 14],
[68, 4, 60, 83, 98],
[14, 30, 14, 63, 65],
[96, 69, 67, 13, 20]],
[[99, 12, 17, 75, 78],
[18, 7, 34, 0, 89],
[33, 61, 32, 59, 99],
[91, 73, 18, 38, 86]]])
n.shape
(3, 4, 5)
image = np.random.randint(0, 255, size=(456, 730, 3))
image.shape
(456, 730, 3)
plt.imshow(image)
<matplotlib.image.AxesImage at 0x7fccb0cac0f0>
8) np.random.randn(d0, d1, ..., dn)
标准正太分布
# 平均值为0, 方差为1的正态分布, 就是标准正态分布.
n = np.random.randn(1000, 30, 2)
n
array([[[ 0.4896418 , -0.60883984],
[ 0.26853052, -0.60839539],
[ 0.31294487, 1.31865127],
...,
[ 0.26797445, 0.11520259],
[ 2.04995086, 0.09461803],
[ 1.07127391, -1.31067053]],
[[ 0.34593488, 0.24465173],
[-0.4157803 , 0.29663043],
[-0.14079407, -0.41392739],
...,
[ 0.90807284, 1.17782632],
[ 0.12260718, -1.16451586],
[ 0.06554271, -1.3787287 ]],
[[ 0.80285065, 0.08647407],
[ 1.33162021, -1.02656728],
[-1.67802604, -1.04042113],
...,
[-0.91478354, 2.76745611],
[-0.6926192 , 0.57983548],
[-0.79930956, -0.57962272]],
...,
[[ 0.20735501, 1.90684661],
[-1.27024817, -0.41497321],
[-1.02678114, -0.49408855],
...,
[ 0.82887736, 0.74195174],
[-0.28360596, 0.85135051],
[-1.50978638, -1.88802704]],
[[-0.02460612, 0.74332003],
[ 0.67222545, -1.60336698],
[ 0.71210037, 0.87587366],
...,
[-1.16626156, -0.1063151 ],
[-0.53933263, 1.2383391 ],
[ 0.33725789, -0.21701002]],
[[ 0.74405161, 1.06175663],
[-1.22704475, 0.261197 ],
[ 0.25032738, -0.10785808],
...,
[ 0.22590857, 2.22273275],
[-0.72196512, -0.07164419],
[-0.58561601, 1.46280227]]])
9)np.random.normal(loc=0.0, scale=1.0, size=None)
import numpy as np
# 正态分布,也叫高斯分布
# 概率密度曲线
# loc = location
# scale =
n = np.random.normal(loc=10, scale=3, size=(1000, 2))
n
array([[ 2.11633679, 10.37104791],
[12.79906613, 9.32691209],
[ 8.73445973, 6.49449323],
...,
[10.90201134, 6.89778288],
[15.53889102, 6.386347 ],
[12.84620178, 5.08296774]])
n.mean()
10.023737800147947
n.std()
3.04452541108808
10) np.random.random(size=None)
生成0到1的随机数,左闭右开
np.random.random(size=(5, 2, 3)) # dimension 纬度
array([[[0.72040153, 0.09664452, 0.27252567],
[0.06722621, 0.7023822 , 0.84930292]],
[[0.87268163, 0.81145728, 0.87352575],
[0.1631786 , 0.80191163, 0.58542209]],
[[0.60120435, 0.04436833, 0.15913068],
[0.90660888, 0.42009378, 0.27366765]],
[[0.00406047, 0.50875819, 0.09918324],
[0.87496182, 0.24859695, 0.97042554]],
[[0.3902245 , 0.73079414, 0.93259252],
[0.96045494, 0.7478134 , 0.25658827]]])
# 和np.random.random一样
np.random.rand(5, 2, 3)
array([[[0.17731902, 0.44709539, 0.34512151],
[0.0648051 , 0.76571718, 0.80429627]],
[[0.61488974, 0.79271908, 0.79005958],
[0.65512793, 0.5050555 , 0.09845543]],
[[0.30139638, 0.36260231, 0.50582467],
[0.75408696, 0.13440877, 0.01720899]],
[[0.41123981, 0.1544165 , 0.52926244],
[0.76862384, 0.35174404, 0.45054936]],
[[0.93168058, 0.05395682, 0.59795456],
[0.77248911, 0.62356132, 0.62255185]]])
二、ndarray的属性
四个必记参数:
ndim: 纬度
shape:形状(各纬度的长度)
size:总长度
dtype:元素类型
n = np.random.randint(0, 100, size=(3, 4, 5))
n
array([[[97, 96, 97, 27, 7],
[63, 74, 32, 72, 96],
[45, 53, 69, 47, 38],
[66, 32, 34, 53, 9]],
[[96, 52, 42, 93, 47],
[99, 81, 33, 36, 27],
[40, 38, 34, 40, 77],
[88, 16, 95, 98, 49]],
[[ 6, 2, 54, 32, 2],
[23, 37, 59, 4, 53],
[58, 22, 61, 50, 24],
[72, 94, 75, 39, 16]]])
n.ndim
3
n.shape
(3, 4, 5)
n.size
60
n.dtype
dtype('int64')
三、ndarray的基本操作
1. 索引
一维与列表完全一致
多维时同理
l = [5, 2, 3, 4, 7]
n = np.array(l)
n
array([5, 2, 3, 4, 7])
n[0]
5
l[0]
5
n2 = np.random.randint(0, 100, size=(4, 5))
n2
array([[73, 64, 29, 35, 61],
[91, 30, 63, 40, 21],
[22, 69, 83, 9, 42],
[13, 34, 0, 71, 30]])
n2[0]
array([73, 64, 29, 35, 61])
n2[0][1]
64
n2[0, 1]
64
n2[1, 3]
40
n3 = np.random.randint(0, 50, size=(3, 4, 3, 6))
n3
array([[[[36, 12, 19, 20, 13, 47],
[ 6, 26, 47, 29, 7, 3],
[14, 46, 23, 13, 23, 1]],
[[13, 33, 14, 49, 14, 28],
[49, 26, 6, 28, 1, 5],
[ 0, 11, 19, 12, 12, 24]],
[[28, 38, 21, 10, 33, 23],
[ 5, 22, 36, 14, 43, 45],
[30, 38, 7, 39, 8, 17]],
[[13, 22, 19, 0, 13, 34],
[13, 31, 35, 30, 38, 5],
[36, 44, 2, 1, 49, 36]]],
[[[10, 28, 22, 49, 4, 24],
[36, 32, 11, 24, 3, 13],
[ 2, 1, 23, 41, 11, 30]],
[[13, 24, 46, 36, 10, 19],
[ 2, 29, 9, 11, 35, 7],
[ 2, 11, 16, 33, 19, 11]],
[[ 4, 17, 45, 44, 36, 23],
[ 4, 21, 49, 33, 12, 33],
[33, 21, 12, 9, 31, 43]],
[[12, 0, 27, 12, 24, 24],
[49, 5, 18, 45, 26, 35],
[19, 7, 11, 1, 2, 24]]],
[[[19, 4, 25, 48, 23, 4],
[24, 12, 48, 3, 46, 14],
[33, 42, 2, 15, 1, 26]],
[[24, 25, 41, 23, 31, 34],
[45, 3, 47, 27, 32, 29],
[47, 5, 34, 11, 49, 41]],
[[42, 36, 11, 24, 46, 22],
[18, 25, 18, 26, 0, 16],
[34, 9, 31, 11, 14, 48]],
[[48, 26, 49, 36, 39, 4],
[ 3, 27, 17, 13, 41, 5],
[31, 8, 29, 15, 33, 45]]]])
n3.shape
(3, 4, 3, 6)
n3.ndim
4
n3[0, 1, 0, 1]
33
# 根据索引修改数据
n3[0, 1, 0, 1] = 88
n3[0, 1, 0, 1]
88
2. 切片
一维与列表完全一致
多维时同理
l = [5, 2, 3, 8, 4]
n = np.array(l)
n
array([5, 2, 3, 8, 4])
l[1:]
[2, 3, 8, 4]
n[1:]
array([2, 3, 8, 4])
# 将列表反转
l[::-1]
[4, 8, 3, 2, 5]
# 将array反转
n[::-1]
array([4, 8, 3, 2, 5])
n2 = np.random.randint(0, 255, size=(4, 5))
n2
array([[129, 125, 60, 10, 114],
[ 90, 209, 119, 37, 66],
[218, 247, 30, 111, 211],
[ 65, 165, 15, 247, 23]])
# 二维array反转
n2[::-1]
array([[ 65, 165, 15, 247, 23],
[218, 247, 30, 111, 211],
[ 90, 209, 119, 37, 66],
[129, 125, 60, 10, 114]])
n2[0:2]
array([[129, 125, 60, 10, 114],
[ 90, 209, 119, 37, 66]])
n2[0:2][0][0:3]
array([129, 125, 60])
3. 变形
使用reshape函数,注意参数是一个tuple!
n = np.random.randint(0, 100, size=(3, 4, 5))
n
array([[[17, 80, 74, 62, 53],
[26, 85, 4, 13, 61],
[73, 84, 14, 97, 99],
[61, 54, 9, 54, 35]],
[[31, 1, 68, 4, 67],
[21, 71, 27, 47, 97],
[19, 0, 65, 48, 2],
[58, 6, 1, 19, 49]],
[[60, 98, 27, 76, 51],
[25, 3, 49, 30, 32],
[74, 89, 29, 26, 71],
[13, 5, 14, 24, 2]]])
n.size
60
n.reshape(6, 10) # 新的array的size必须和原array的size相同
array([[17, 80, 74, 62, 53, 26, 85, 4, 13, 61],
[73, 84, 14, 97, 99, 61, 54, 9, 54, 35],
[31, 1, 68, 4, 67, 21, 71, 27, 47, 97],
[19, 0, 65, 48, 2, 58, 6, 1, 19, 49],
[60, 98, 27, 76, 51, 25, 3, 49, 30, 32],
[74, 89, 29, 26, 71, 13, 5, 14, 24, 2]])
n.size # 新的array的size必须和原array的size相同
60
# # reshape之后的size不能发生变化.
n.reshape(5, 2, 6)
array([[[17, 80, 74, 62, 53, 26],
[85, 4, 13, 61, 73, 84]],
[[14, 97, 99, 61, 54, 9],
[54, 35, 31, 1, 68, 4]],
[[67, 21, 71, 27, 47, 97],
[19, 0, 65, 48, 2, 58]],
[[ 6, 1, 19, 49, 60, 98],
[27, 76, 51, 25, 3, 49]],
[[30, 32, 74, 89, 29, 26],
[71, 13, 5, 14, 24, 2]]])
np.reshape(n, (4, 15))
array([[17, 80, 74, 62, 53, 26, 85, 4, 13, 61, 73, 84, 14, 97, 99],
[61, 54, 9, 54, 35, 31, 1, 68, 4, 67, 21, 71, 27, 47, 97],
[19, 0, 65, 48, 2, 58, 6, 1, 19, 49, 60, 98, 27, 76, 51],
[25, 3, 49, 30, 32, 74, 89, 29, 26, 71, 13, 5, 14, 24, 2]])
4. 级联
- np.concatenate()
- 级联的参数是列表:一定要加中括号或小括号
- 维度必须相同
- 形状相符
- 【重点】级联的方向默认是shape这个tuple的第一个值所代表的维度方向
- 可通过axis参数改变级联的方向
n = np.random.randint(0, 255, size=(5, 6))
n2 = np.random.randint(0, 244, size=(5, 6))
display(n, n2)
array([[156, 180, 226, 253, 245, 90],
[154, 46, 81, 158, 170, 50],
[231, 71, 17, 95, 205, 111],
[ 24, 104, 198, 2, 2, 7],
[ 29, 189, 33, 206, 55, 148]])
array([[178, 163, 174, 28, 47, 144],
[ 85, 138, 120, 2, 5, 13],
[ 60, 53, 24, 60, 243, 227],
[ 26, 156, 22, 146, 105, 121],
[215, 130, 160, 161, 105, 49]])
# axis=0 , 增加了行数, 垂直方向级联
np.concatenate((n, n2), axis=0)
array([[156, 180, 226, 253, 245, 90],
[154, 46, 81, 158, 170, 50],
[231, 71, 17, 95, 205, 111],
[ 24, 104, 198, 2, 2, 7],
[ 29, 189, 33, 206, 55, 148],
[178, 163, 174, 28, 47, 144],
[ 85, 138, 120, 2, 5, 13],
[ 60, 53, 24, 60, 243, 227],
[ 26, 156, 22, 146, 105, 121],
[215, 130, 160, 161, 105, 49]])
# axis=1, 增加列数, 在水平方向上级联
np.concatenate((n, n2), axis=1)
array([[156, 180, 226, 253, 245, 90, 178, 163, 174, 28, 47, 144],
[154, 46, 81, 158, 170, 50, 85, 138, 120, 2, 5, 13],
[231, 71, 17, 95, 205, 111, 60, 53, 24, 60, 243, 227],
[ 24, 104, 198, 2, 2, 7, 26, 156, 22, 146, 105, 121],
[ 29, 189, 33, 206, 55, 148, 215, 130, 160, 161, 105, 49]])
- np.hstack与np.vstack
- 水平级联与垂直级联,处理自己,进行维度的变更
array级联
n1 = np.random.randint(0, 100, size=(3, 4))
n2 = np.random.randint(0, 100, size=(3, 4))
display(n1, n2)
array([[46, 1, 47, 3],
[54, 84, 77, 27],
[52, 99, 88, 25]])
array([[10, 95, 45, 64],
[58, 67, 53, 86],
[63, 70, 58, 89]])
np.hstack((n1, n2)) # h = horizontal
array([[46, 1, 47, 3, 10, 95, 45, 64],
[54, 84, 77, 27, 58, 67, 53, 86],
[52, 99, 88, 25, 63, 70, 58, 89]])
np.vstack((n1, n2)) # v = vertical
array([[46, 1, 47, 3],
[54, 84, 77, 27],
[52, 99, 88, 25],
[10, 95, 45, 64],
[58, 67, 53, 86],
[63, 70, 58, 89]])
图片级联
import matplotlib.pyplot as plt
cat = plt.imread('cat.jpg')
cat.shape
(456, 730, 3)
type(cat)
numpy.ndarray
plt.imshow(cat)
<matplotlib.image.AxesImage at 0x7eff3ddb4438>
image = np.random.randint(0, 255, size=(456, 730, 3))
image.shape
(456, 730, 3)
plt.imshow(image)
<matplotlib.image.AxesImage at 0x7eff3d901d30>
new_image = np.hstack((cat, image))
plt.imshow(new_image)
<matplotlib.image.AxesImage at 0x7eff3c07dc88>
new_image = np.vstack((cat, image))
plt.imshow(new_image)
<matplotlib.image.AxesImage at 0x7eff368f9278>
5. 切分
与级联类似,三个函数完成切分工作:
- np.split
- np.vsplit
- np.hsplit
n = np.random.randint(0, 100, size=(4, 5))
n
array([[82, 42, 91, 36, 93],
[ 5, 83, 45, 66, 51],
[13, 18, 5, 56, 5],
[10, 48, 62, 47, 27]])
# 左闭右开区间
np.split(n, [1, 2])
[array([[82, 42, 91, 36, 93]]),
array([[ 5, 83, 45, 66, 51]]),
array([[13, 18, 5, 56, 5],
[10, 48, 62, 47, 27]])]
np.split(n, (2, 2))
[array([[82, 42, 91, 36, 93],
[ 5, 83, 45, 66, 51]]),
array([], shape=(0, 5), dtype=int64),
array([[13, 18, 5, 56, 5],
[10, 48, 62, 47, 27]])]
n2 = np.random.randint(0, 100, size=(7, 5))
n2
array([[99, 27, 16, 63, 93],
[85, 65, 7, 37, 25],
[97, 76, 94, 87, 75],
[93, 93, 26, 63, 62],
[ 6, 35, 7, 45, 22],
[12, 9, 6, 75, 8],
[ 6, 66, 72, 94, 58]])
np.split(n2, (2, 5))
[array([[99, 27, 16, 63, 93],
[85, 65, 7, 37, 25]]), array([[97, 76, 94, 87, 75],
[93, 93, 26, 63, 62],
[ 6, 35, 7, 45, 22]]), array([[12, 9, 6, 75, 8],
[ 6, 66, 72, 94, 58]])]
np.split(n2, (2,), axis=1)
[array([[99, 27],
[85, 65],
[97, 76],
[93, 93],
[ 6, 35],
[12, 9],
[ 6, 66]]), array([[16, 63, 93],
[ 7, 37, 25],
[94, 87, 75],
[26, 63, 62],
[ 7, 45, 22],
[ 6, 75, 8],
[72, 94, 58]])]
cat = plt.imread('cat.jpg')
cat.shape
(456, 730, 3)
plt.imshow(cat)
<matplotlib.image.AxesImage at 0x7eff3686aa20>
res = np.split(cat, 2, axis=0)
res
[array([[[231, 186, 131],
[232, 187, 132],
[233, 188, 133],
...,
[100, 54, 54],
[ 92, 48, 47],
[ 85, 43, 44]],
[[232, 187, 132],
[232, 187, 132],
[233, 188, 133],
...,
[100, 54, 54],
[ 92, 48, 47],
[ 84, 42, 43]],
[[232, 187, 132],
[233, 188, 133],
[233, 188, 133],
...,
[ 99, 53, 53],
[ 91, 47, 46],
[ 83, 41, 42]],
...,
[[181, 110, 92],
[182, 111, 93],
[184, 111, 94],
...,
[ 94, 46, 44],
[ 89, 43, 43],
[ 89, 43, 43]],
[[179, 108, 90],
[181, 110, 92],
[183, 110, 93],
...,
[ 94, 46, 44],
[ 91, 43, 43],
[ 89, 43, 43]],
[[179, 108, 90],
[181, 110, 92],
[182, 111, 93],
...,
[ 95, 45, 44],
[ 91, 43, 43],
[ 89, 43, 43]]], dtype=uint8), array([[[178, 110, 91],
[181, 110, 92],
[182, 111, 93],
...,
[ 95, 45, 44],
[ 91, 43, 43],
[ 91, 43, 43]],
[[178, 110, 91],
[180, 112, 93],
[182, 111, 93],
...,
[ 95, 45, 44],
[ 91, 41, 42],
[ 90, 42, 42]],
[[176, 109, 90],
[179, 111, 92],
[180, 112, 93],
...,
[ 96, 45, 44],
[ 92, 42, 43],
[ 91, 43, 43]],
...,
[[199, 119, 82],
[199, 119, 82],
[200, 120, 83],
...,
[189, 99, 65],
[187, 97, 63],
[187, 97, 63]],
[[199, 119, 82],
[199, 119, 82],
[199, 119, 82],
...,
[188, 98, 64],
[186, 96, 62],
[188, 95, 62]],
[[199, 119, 82],
[199, 119, 82],
[199, 119, 82],
...,
[188, 98, 64],
[188, 95, 62],
[188, 95, 62]]], dtype=uint8)]
plt.imshow(res[0])
<matplotlib.image.AxesImage at 0x7eff3653c828>
res = np.split(cat, 2, axis=1)
len(res)
2
type(res)
list
plt.imshow(res[0])
<matplotlib.image.AxesImage at 0x7eff36478588>
plt.imshow(res[1])
<matplotlib.image.AxesImage at 0x7eff364466a0>
6. 副本
所有赋值运算不会为ndarray的任何元素创建副本。对赋值后的对象的操作也对原来的对象生效。
l = [1, 2, 3, 5]
l2 = l
l2[2] = 8
l
[1, 2, 8, 5]
import numpy as np
n1 = np.arange(10)
n2 = n1
n2[3] = 100
n1
array([ 0, 1, 2, 100, 4, 5, 6, 7, 8, 9])
可以使用copy()函数创建副本
n3 = np.random.randint(0, 20, size=(3, 5))
n3
array([[13, 11, 19, 8, 9],
[ 2, 5, 17, 19, 11],
[ 4, 5, 11, 0, 15]])
n4 = n3.copy()
n4[0, 1] = 88
n4
array([[13, 88, 19, 8, 9],
[ 2, 5, 17, 19, 11],
[ 4, 5, 11, 0, 15]])
n3
array([[13, 11, 19, 8, 9],
[ 2, 5, 17, 19, 11],
[ 4, 5, 11, 0, 15]])
四、ndarray的聚合操作
1. 求和np.sum
n = np.arange(11)
n
array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10])
np.sum(n)
55
n2 = np.random.randint(0, 100, size=(3, 4))
n2
array([[89, 44, 29, 66],
[ 2, 4, 56, 95],
[86, 41, 2, 90]])
np.sum(n2, axis=1)
array([228, 157, 219])
2. 最大最小值:np.max/ np.min
同理
n = np.arange(11)
n
array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10])
np.median(n)
5.0
np.mean(n)
5.0
n2 = np.random.randint(0, 100, size=10)
n2
array([63, 26, 83, 31, 2, 68, 38, 76, 96, 4])
np.mean(n2)
48.7
np.median(n2)
50.5
np.max(n2)
96
np.min(n2)
2
n3 = np.random.randint(0, 100, size=(5, 6))
n3
array([[ 4, 12, 9, 53, 51, 4],
[24, 72, 8, 59, 42, 81],
[58, 38, 2, 81, 34, 74],
[70, 89, 11, 50, 66, 86],
[ 2, 77, 86, 11, 84, 59]])
np.max(n3, axis=0)
array([70, 89, 86, 81, 84, 86])
np.max(n3, axis=1)
array([53, 81, 81, 89, 86])
3. 其他聚合操作
Function Name | NaN-safe Version | Description |
---|---|---|
np.sum | np.nansum | Compute sum of elements |
np.prod | np.nanprod | Compute product of elements |
np.mean | np.nanmean | Compute mean of elements |
np.std | np.nanstd | Compute standard deviation |
np.var | np.nanvar | Compute variance |
np.min | np.nanmin | Find minimum value |
np.max | np.nanmax | Find maximum value |
np.argmin | np.nanargmin | Find index of minimum value |
np.argmax | np.nanargmax | Find index of maximum value |
np.median | np.nanmedian | Compute median of elements |
np.percentile | np.nanpercentile | Compute rank-based statistics of elements |
np.any | N/A | Evaluate whether any elements are true |
np.all | N/A | Evaluate whether all elements are true |
np.power | 幂运算 |
n = np.random.randint(0, 100, size=(5, 6))
n
array([[35, 87, 31, 31, 62, 48],
[19, 15, 25, 39, 31, 23],
[43, 61, 58, 49, 30, 80],
[83, 48, 94, 38, 10, 9],
[ 9, 11, 1, 28, 31, 70]])
np.argmin(n, axis=0)
array([4, 4, 4, 4, 3, 3])
n2 = np.random.randint(0, 10, size=(2, 5))
n2
array([[1, 5, 1, 8, 4],
[1, 6, 9, 8, 5]])
np.reshape(n2, (-1,))
array([1, 5, 1, 8, 4, 1, 6, 9, 8, 5])
np.sum 和 np.nansum 的区别
nan not a number
a = np.array([1, 2, np.nan])
a
array([ 1., 2., nan])
np.sum(a)
nan
np.nansum(a)
3.0
操作文件
使用pandas打开文件president_heights.csv
获取文件中的数据
!pip3 install pandas
import pandas as pd
data = pd.read_csv('president_heights.csv')
type(data)
pandas.core.frame.DataFrame
data
order | name | height(cm) | |
---|---|---|---|
0 | 1 | George Washington | 189 |
1 | 2 | John Adams | 170 |
2 | 3 | Thomas Jefferson | 189 |
3 | 4 | James Madison | 163 |
4 | 5 | James Monroe | 183 |
5 | 6 | John Quincy Adams | 171 |
6 | 7 | Andrew Jackson | 185 |
7 | 8 | Martin Van Buren | 168 |
8 | 9 | William Henry Harrison | 173 |
9 | 10 | John Tyler | 183 |
10 | 11 | James K. Polk | 173 |
11 | 12 | Zachary Taylor | 173 |
12 | 13 | Millard Fillmore | 175 |
13 | 14 | Franklin Pierce | 178 |
14 | 15 | James Buchanan | 183 |
15 | 16 | Abraham Lincoln | 193 |
16 | 17 | Andrew Johnson | 178 |
17 | 18 | Ulysses S. Grant | 173 |
18 | 19 | Rutherford B. Hayes | 174 |
19 | 20 | James A. Garfield | 183 |
20 | 21 | Chester A. Arthur | 183 |
21 | 23 | Benjamin Harrison | 168 |
22 | 25 | William McKinley | 170 |
23 | 26 | Theodore Roosevelt | 178 |
24 | 27 | William Howard Taft | 182 |
25 | 28 | Woodrow Wilson | 180 |
26 | 29 | Warren G. Harding | 183 |
27 | 30 | Calvin Coolidge | 178 |
28 | 31 | Herbert Hoover | 182 |
29 | 32 | Franklin D. Roosevelt | 188 |
30 | 33 | Harry S. Truman | 175 |
31 | 34 | Dwight D. Eisenhower | 179 |
32 | 35 | John F. Kennedy | 183 |
33 | 36 | Lyndon B. Johnson | 193 |
34 | 37 | Richard Nixon | 182 |
35 | 38 | Gerald Ford | 183 |
36 | 39 | Jimmy Carter | 177 |
37 | 40 | Ronald Reagan | 185 |
38 | 41 | George H. W. Bush | 188 |
39 | 42 | Bill Clinton | 188 |
40 | 43 | George W. Bush | 182 |
41 | 44 | Barack Obama | 185 |
heights = data['height(cm)']
heights
0 189
1 170
2 189
3 163
4 183
5 171
6 185
7 168
8 173
9 183
10 173
11 173
12 175
13 178
14 183
15 193
16 178
17 173
18 174
19 183
20 183
21 168
22 170
23 178
24 182
25 180
26 183
27 178
28 182
29 188
30 175
31 179
32 183
33 193
34 182
35 183
36 177
37 185
38 188
39 188
40 182
41 185
Name: height(cm), dtype: int64
type(heights)
pandas.core.series.Series
import numpy as np
np.max(heights)
193
np.mean(heights)
179.73809523809524
np.std(heights)
6.931843442745892
五、ndarray的矩阵操作
1. 基本矩阵操作
1) 算术运算符:
- 加减乘除
n = np.random.randint(0, 10, size=(4, 5))
n
array([[8, 0, 0, 9, 3],
[5, 4, 3, 1, 0],
[1, 8, 1, 2, 9],
[7, 5, 0, 0, 9]])
# 加
n + 1
array([[ 9, 1, 1, 10, 4],
[ 6, 5, 4, 2, 1],
[ 2, 9, 2, 3, 10],
[ 8, 6, 1, 1, 10]])
# 减
n - 1
array([[ 7, -1, -1, 8, 2],
[ 4, 3, 2, 0, -1],
[ 0, 7, 0, 1, 8],
[ 6, 4, -1, -1, 8]])
n2 = np.random.randint(0, 10, size=(4, 5))
n2
array([[2, 4, 9, 2, 9],
[8, 8, 0, 6, 1],
[8, 8, 7, 9, 2],
[4, 2, 3, 0, 3]])
n + n2
array([[10, 4, 9, 11, 12],
[13, 12, 3, 7, 1],
[ 9, 16, 8, 11, 11],
[11, 7, 3, 0, 12]])
n3 = np.random.randint(0, 10, size=(4, 5))
n3
array([[6, 8, 6, 4, 8],
[3, 4, 7, 1, 1],
[7, 9, 4, 6, 2],
[3, 3, 6, 2, 5]])
n2 + n3
array([[ 8, 12, 15, 6, 17],
[11, 12, 7, 7, 2],
[15, 17, 11, 15, 4],
[ 7, 5, 9, 2, 8]])
2) 矩阵积np.dot()
n1 = np.random.randint(0, 10, size=(2, 3))
n1
array([[5, 3, 6],
[9, 4, 4]])
n2 = np.random.randint(0, 10, size=(3, 4))
n2
array([[6, 5, 8, 8],
[0, 7, 6, 2],
[3, 3, 7, 4]])
np.dot(n1, n2)
array([[ 48, 64, 100, 70],
[ 66, 85, 124, 96]])
2. 广播机制
【重要】ndarray广播机制的两条规则
- 规则一:为缺失的维度补1
- 规则二:假定缺失元素用已有值填充
例1:
m = np.ones((2, 3))
a = np.arange(3)
求M+a
m = np.ones((2, 3), dtype=int)
m
array([[1, 1, 1],
[1, 1, 1]])
n = np.arange(3)
n
array([0, 1, 2])
m + n
array([[1, 2, 3],
[1, 2, 3]])
例2:
a = np.arange(3).reshape((3, 1))
b = np.arange(3)
求a+b
a = np.arange(3).reshape((3, 1))
a
array([[0],
[1],
[2]])
b = np.arange(3)
b
array([0, 1, 2])
a + b
array([[0, 1, 2],
[1, 2, 3],
[2, 3, 4]])
习题
a = np.ones((4, 1))
b = np.arange(4)
求a+b
a = np.ones((4, 1))
a
array([[1.],
[1.],
[1.],
[1.]])
b = np.arange(4)
b
array([0, 1, 2, 3])
a + b
array([[1., 2., 3., 4.],
[1., 2., 3., 4.],
[1., 2., 3., 4.],
[1., 2., 3., 4.]])
六、ndarray的排序
小测验:
使用以上所学numpy的知识,对一个ndarray对象进行选择排序。
def Sortn(x):
代码越短越好
n = [5, 2, 3, 6, 9]
def bubble(n):
for i in range(len(n) - 1):
for j in range(i+1, len(n)):
if n[i] > n[j]:
n[i], n[j] = n[j], n[i]
bubble(n)
n
[2, 3, 5, 6, 9]
# 选择排序
def select(n):
for i in range(len(n)):
# 选出最小直的索引
index = np.argmin(n[i:]) + i
# 把最小值和当前值的位置换一下
n[i], n[index] = n[index], n[i]
n = [4, 6, 1, 0, 3]
select(n)
n
[0, 1, 3, 4, 6]
1. 快速排序
np.sort()与ndarray.sort()都可以,但有区别:
- np.sort()不改变输入
- ndarray.sort()本地处理,不占用空间,但改变输入
# 排序只对1维的数据有意义
n = np.arange(0, 10)
n
array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
np.random.shuffle(n)
n
array([7, 1, 3, 9, 0, 8, 6, 5, 2, 4])
# 第一种np.sort()
np.sort(n)
array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
n
array([7, 1, 3, 9, 0, 8, 6, 5, 2, 4])
# 对象的sort方法
n.sort()
n
array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
部分排序
n = np.arange(0, 100)
np.random.shuffle(n)
n
array([76, 23, 49, 39, 17, 79, 9, 91, 56, 84, 40, 11, 4, 83, 7, 60, 25,
75, 88, 26, 99, 24, 43, 85, 63, 59, 68, 22, 71, 51, 34, 86, 21, 36,
13, 20, 10, 45, 37, 81, 53, 69, 78, 14, 92, 41, 65, 74, 87, 95, 44,
72, 96, 35, 3, 46, 73, 38, 0, 16, 97, 64, 62, 89, 28, 55, 19, 5,
18, 48, 2, 54, 94, 32, 98, 8, 27, 52, 67, 29, 93, 82, 58, 77, 1,
33, 47, 42, 90, 31, 70, 30, 50, 57, 12, 6, 15, 61, 66, 80])
# 找出前五个最大的数
np.partition(n, kth=-5)
array([58, 76, 49, 39, 17, 79, 9, 66, 56, 61, 40, 11, 4, 15, 7, 60, 25,
75, 6, 26, 12, 24, 43, 57, 63, 59, 68, 22, 71, 51, 34, 50, 21, 36,
13, 20, 10, 45, 37, 30, 53, 69, 78, 14, 70, 41, 65, 74, 31, 23, 44,
72, 42, 35, 3, 46, 73, 38, 0, 16, 47, 64, 62, 33, 28, 55, 19, 5,
18, 48, 2, 54, 1, 32, 77, 8, 27, 52, 67, 29, 80, 86, 82, 91, 84,
89, 83, 88, 90, 87, 85, 81, 92, 94, 93, 95, 99, 96, 98, 97])
# 找出最小的五个数
np.partition(n, kth=5)
array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 11, 9, 13, 29, 28, 25,
16, 15, 26, 19, 24, 18, 17, 23, 14, 27, 22, 20, 21, 30, 50, 34, 36,
51, 71, 68, 45, 37, 76, 53, 69, 78, 59, 70, 41, 65, 74, 31, 63, 44,
72, 42, 35, 57, 46, 73, 38, 43, 75, 47, 64, 62, 33, 60, 55, 40, 61,
56, 48, 66, 54, 79, 32, 77, 39, 49, 52, 67, 58, 80, 82, 93, 98, 94,
89, 97, 96, 90, 87, 92, 81, 86, 85, 99, 88, 83, 84, 91, 95])