numpy基础用法学习
numpy get started
导入numpy库,并查看numpy版本
import numpy as np
np.__version__
'1.14.0'
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
data = [1, 2, 3]
nd = np.array(data)
nd
array([1, 2, 3])
type(nd)
#ndarray 这样的数据
#查看整体的类型
numpy.ndarray
type(data)
list
nd.dtype
#查看数据的类型
dtype('int32')
nd1 = np.array([1,2,3.2])
nd1.dtype
#统一的原则 int < float < string
dtype('float64')
nd2 = np.array([1,2,3.4, "qwe"])
nd2.dtype
dtype('<U32')
nd3 = np.array([[1,2],[3,4]])
nd3
array([[1, 2],
[3, 4]])
nd4 = np.array([[[1,2],[3,4]],[[1,2],[3,4]]])
nd4
array([[[1, 2],
[3, 4]],
[[1, 2],
[3, 4]]])
nd4.shape
(2, 2, 2)
#扩展
import matplotlib.pyplot as plt
cat = plt.imread("./cat.jpg")
type(cat)
numpy.ndarray
plt.imshow(cat[:300,:200])
plt.show()
cat.shape
#查看形状的 rgb jpg 0-255
(456, 730, 3)
cat
#三维的数据
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)
2. 使用np的routines函数创建
包含以下常见创建方法:
- np.ones(shape, dtype=None, order='C')
ones = np.ones((456,730,3), dtype = "float")
#shape 形状, 元祖
ones
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., 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., 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.],
[1., 1., 1.],
...,
[1., 1., 1.],
[1., 1., 1.],
[1., 1., 1.]]])
plt.imshow(ones)
plt.show()
#0-1 0 代表黑色的 1 白色的 png格式的图片
#切片赋值
ones[::,::,1:] = 0
ones
array([[[1., 0., 0.],
[1., 0., 0.],
[1., 0., 0.],
...,
[1., 0., 0.],
[1., 0., 0.],
[1., 0., 0.]],
[[1., 0., 0.],
[1., 0., 0.],
[1., 0., 0.],
...,
[1., 0., 0.],
[1., 0., 0.],
[1., 0., 0.]],
[[1., 0., 0.],
[1., 0., 0.],
[1., 0., 0.],
...,
[1., 0., 0.],
[1., 0., 0.],
[1., 0., 0.]],
...,
[[1., 0., 0.],
[1., 0., 0.],
[1., 0., 0.],
...,
[1., 0., 0.],
[1., 0., 0.],
[1., 0., 0.]],
[[1., 0., 0.],
[1., 0., 0.],
[1., 0., 0.],
...,
[1., 0., 0.],
[1., 0., 0.],
[1., 0., 0.]],
[[1., 0., 0.],
[1., 0., 0.],
[1., 0., 0.],
...,
[1., 0., 0.],
[1., 0., 0.],
[1., 0., 0.]]])
plt.imshow(ones)
plt.show()
#1 0 0
ones[::,::,0] = 0.3
ones
array([[[0.3, 0. , 0. ],
[0.3, 0. , 0. ],
[0.3, 0. , 0. ],
...,
[0.3, 0. , 0. ],
[0.3, 0. , 0. ],
[0.3, 0. , 0. ]],
[[0.3, 0. , 0. ],
[0.3, 0. , 0. ],
[0.3, 0. , 0. ],
...,
[0.3, 0. , 0. ],
[0.3, 0. , 0. ],
[0.3, 0. , 0. ]],
[[0.3, 0. , 0. ],
[0.3, 0. , 0. ],
[0.3, 0. , 0. ],
...,
[0.3, 0. , 0. ],
[0.3, 0. , 0. ],
[0.3, 0. , 0. ]],
...,
[[0.3, 0. , 0. ],
[0.3, 0. , 0. ],
[0.3, 0. , 0. ],
...,
[0.3, 0. , 0. ],
[0.3, 0. , 0. ],
[0.3, 0. , 0. ]],
[[0.3, 0. , 0. ],
[0.3, 0. , 0. ],
[0.3, 0. , 0. ],
...,
[0.3, 0. , 0. ],
[0.3, 0. , 0. ],
[0.3, 0. , 0. ]],
[[0.3, 0. , 0. ],
[0.3, 0. , 0. ],
[0.3, 0. , 0. ],
...,
[0.3, 0. , 0. ],
[0.3, 0. , 0. ],
[0.3, 0. , 0. ]]])
plt.imshow(ones)
plt.show()
- np.zeros(shape, dtype=float, order='C')
zeros = np.zeros((456,730,3), dtype = "float")
zeros
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., 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., 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.],
[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., 0., 0.],
[0., 0., 0.],
[0., 0., 0.]]])
plt.imshow(zeros)
<matplotlib.image.AxesImage at 0x9992e80>
- np.full(shape, fill_value, dtype=None, order='C')
nd4 = np.full(12, fill_value=1024)
nd4
array([1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024,
1024])
#变形,注意 你要变形的数据有多少? 不能超过变形总的长度
#cannot reshape array of size 12 into shape (3,5)
nd5 = nd4.reshape((3,5))
nd5
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<ipython-input-40-0baec20d7413> in <module>()
1 #变形
----> 2 nd5 = nd4.reshape((3,5))
3 nd5
ValueError: cannot reshape array of size 12 into shape (3,5)
nd5 = nd4.reshape((1,12))
nd5
#reshape在咱们以后的学习经常使用
array([[1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024,
1024]])
#扩展
cat
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)
cat.shape
(456, 730, 3)
#需求:把猫的图片颠倒一下
cat2 = cat[::-1,::-1,::]
plt.imshow(cat2)
<matplotlib.image.AxesImage at 0x99d7eb8>
- np.eye(N, M=None, k=0, dtype=float)
对角线为1其他的位置为0
#产生一个 单元矩阵
np.eye(5)
array([[1., 0., 0., 0., 0.],
[0., 1., 0., 0., 0.],
[0., 0., 1., 0., 0.],
[0., 0., 0., 1., 0.],
[0., 0., 0., 0., 1.]])
#每一个有解的矩阵 到最后都可以化成单元矩阵 满秩矩阵
2x + 3y + 4z = 12
4x + 7y + 8z = 32
2 3 4
4 7 8 => 最后化成单元矩阵 是没办法化成的单元矩阵
4 6 8
- np.linspace(start, stop, num=50, endpoint=True, retstep=False, dtype=None)
#lin linear 线性的
np.linspace(0,10,num = 13)
array([ 0. , 0.83333333, 1.66666667, 2.5 , 3.33333333,
4.16666667, 5. , 5.83333333, 6.66666667, 7.5 ,
8.33333333, 9.16666667, 10. ])
np.log2(4)
2.0
np.logspace(-3,1,2)
array([1.e-03, 1.e+01])
- np.arange([start, ]stop, [step, ]dtype=None)
#我会经常写
np.arange(start = 5, stop = 10, step = 2,dtype = "float")
array([5., 7., 9.])
np.arange(0,150,10)
array([ 0, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 110, 120,
130, 140])
- np.random.randint(low, high=None, size=None, dtype='l')
np.random.randint(0,150,size = 10)
array([ 57, 116, 133, 84, 141, 32, 63, 91, 93, 16])
dog = np.random.randint(0,255,size = (456,730,3))
dog.shape
(456, 730, 3)
cat.dtype
dtype('uint8')
dog.dtype
dtype('int32')
#数据类型转换
dog = dog.astype('uint8')
plt.imshow(dog)
<matplotlib.image.AxesImage at 0xfcca828>
- np.random.randn(d0, d1, ..., dn)
标准正太分布
np.random.randn(10,2,1)
#正太分布会搞出来数据,是两边低,中间高的数据
array([[[ 1.39438673],
[-0.78456615]],
[[-0.59132977],
[ 2.23663625]],
[[ 0.61258477],
[-0.84729158]],
[[ 1.37855508],
[ 1.697815 ]],
[[-0.06004384],
[ 0.98147252]],
[[-1.20190404],
[-0.77774525]],
[[ 1.34400589],
[ 0.23112796]],
[[-0.31579586],
[-0.11644608]],
[[-0.11822406],
[ 0.26001606]],
[[ 0.03766789],
[ 0.80169127]]])
9)np.random.normal(loc=0.0, scale=1.0, size=None)
#只是需要知道他可以搞出来数据
np.random.normal(loc = 175, scale =10,size = 10 )
array([176.45148898, 179.9089715 , 173.65923279, 172.36118888,
169.66272673, 158.76980334, 165.3742424 , 173.52898147,
175.84535943, 183.92875259])
- np.random.random(size=None)
生成0到1的随机数,左闭右开
np.random.random(size = (2,2))
array([[0.64468214, 0.54496107],
[0.20529068, 0.0482465 ]])
使用随机数成成一张图片
二、ndarray的属性
4个必记参数:
ndim:维度
shape:形状(各维度的长度)
size:总长度
dtype:元素类型
cat = plt.imread("./cat.jpg")
cat
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)
cat.ndim
#dimension 维度的意思
3
cat.shape
#这个每天都要用,特别是数据分析
(456, 730, 3)
cat.size
998640
456*730*3
998640
cat.dtype
dtype('uint8')
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 = [1,2,3,4,5,6]
l[-1]
#list
6
nd = np.random.randint(0,100,size = 12)
nd[2]
97
nd = np.random.randint(0,100,size =(4,4))
nd
array([[48, 85, 53, 22],
[24, 36, 26, 31],
[38, 26, 56, 47],
[22, 80, 50, 9]])
nd[0, 1]
#通过索引取出来数据
85
根据索引修改数据
nd[2,2] = 2100
nd
array([[ 48, 85, 53, 22],
[ 24, 36, 26, 31],
[ 38, 26, 2100, 47],
[ 22, 80, 50, 9]])
2. 切片
一维与列表完全一致
多维时同理
nd
array([[ 48, 85, 53, 22],
[ 24, 36, 26, 31],
[ 38, 26, 2100, 47],
[ 22, 80, 50, 9]])
nd[:-2]
array([[48, 85, 53, 22],
[24, 36, 26, 31]])
nd[:-2] = 10
nd
array([[ 10, 10, 10, 10],
[ 10, 10, 10, 10],
[ 38, 26, 2100, 47],
[ 22, 80, 50, 9]])
np.random.randint(0,10,size = (2,1))
array([[2],
[8]])
nd[:2, 1:3] = np.random.randint(0,10,size = (2,1))
nd
#广播机制,如果在赋值的时候,不充分,numpy会自动进行复制
array([[ 10, 0, 0, 10],
[ 10, 8, 8, 10],
[ 38, 26, 2100, 47],
[ 22, 80, 50, 9]])
将数据反转,例如[1,2,3]---->[3,2,1]
nd1 = np.random.randint(0,100,size = 11)
nd1
array([53, 0, 12, 83, 98, 46, 36, 96, 21, 51, 34])
nd1[::-1]
array([34, 51, 21, 96, 36, 46, 98, 83, 12, 0, 53])
#::叫步幅
nd1[::-2]
array([34, 21, 36, 98, 12, 53])
nd
array([[ 10, 0, 0, 10],
[ 10, 8, 8, 10],
[ 38, 26, 2100, 47],
[ 22, 80, 50, 9]])
nd[::1,::-1]
array([[ 10, 0, 0, 10],
[ 10, 8, 8, 10],
[ 47, 2100, 26, 38],
[ 9, 50, 80, 22]])
nd[::-1,::-1]
array([[ 9, 50, 80, 22],
[ 47, 2100, 26, 38],
[ 10, 8, 8, 10],
[ 10, 0, 0, 10]])
两个::进行切片
3. 变形
使用reshape函数,注意参数是一个tuple!
#扩展
cat = plt.imread("./cat.jpg")
cat_f = cat/255.0
plt.imshow(cat_f)
<matplotlib.image.AxesImage at 0xd380b70>
fish = plt.imread("./fish.png")
plt.imshow(fish)
<matplotlib.image.AxesImage at 0xd2b3940>
fish
array([[[0.29411766, 0.39215687, 0.46666667],
[0.46666667, 0.4862745 , 0.49803922],
[0.4627451 , 0.4862745 , 0.5019608 ],
...,
[0.4627451 , 0.48235294, 0.49803922],
[0.45882353, 0.47843137, 0.49803922],
[0.21960784, 0.33333334, 0.44313726]],
[[0.2901961 , 0.3764706 , 0.44313726],
[0.627451 , 0.6156863 , 0.60784316],
[0.85490197, 0.85490197, 0.84705883],
...,
[0.8627451 , 0.85882354, 0.8509804 ],
[0.8509804 , 0.8509804 , 0.84313726],
[0.30588236, 0.42352942, 0.5254902 ]],
[[0.28235295, 0.37254903, 0.4392157 ],
[0.6666667 , 0.6627451 , 0.654902 ],
[1. , 1. , 1. ],
...,
[1. , 1. , 1. ],
[1. , 1. , 1. ],
[0.35686275, 0.4745098 , 0.5764706 ]],
...,
[[0.4509804 , 0.45882353, 0.45882353],
[0.6509804 , 0.6509804 , 0.64705884],
[0.99215686, 0.99215686, 0.9843137 ],
...,
[1. , 0.99607843, 0.9882353 ],
[0.9843137 , 0.9882353 , 0.98039216],
[0.36078432, 0.49019608, 0.6 ]],
[[0.4509804 , 0.45882353, 0.45882353],
[0.6509804 , 0.6509804 , 0.64705884],
[0.99215686, 0.99215686, 0.9843137 ],
...,
[1. , 0.99607843, 0.9882353 ],
[0.9843137 , 0.9882353 , 0.98039216],
[0.36078432, 0.49019608, 0.6 ]],
[[0.44705883, 0.45490196, 0.45490196],
[0.65882355, 0.654902 , 0.654902 ],
[1. , 1. , 1. ],
...,
[1. , 1. , 1. ],
[1. , 1. , 1. ],
[0.36078432, 0.49411765, 0.6 ]]], dtype=float32)
#拿出来鱼头
fish_head = fish[50:175, 50:180]
plt.imshow(fish_head)
<matplotlib.image.AxesImage at 0xd41bd30>
#把猫挑选出来一部分,把鱼头贴上去
cat_f[120:245,220:350] = fish_head
# cat.flags.writeable = True
# fish.flags.writeable = True
plt.imshow(cat_f)
#肯定会报错,一个是png 一个是jpg
<matplotlib.image.AxesImage at 0xd4be9b0>
4. 级联
- np.concatenate()
级联需要注意的点:
- 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
import numpy as np
nd1 = np.random.randint(0,10,size = (4,6))
nd2 = np.random.randint(50,100,size =(2,6))
np.concatenate([nd1, nd2])
#第一个参数 可以传一个list或者tuple 第二个参数是轴 axis
#默认的轴等于0 行上面进行级联
array([[ 3, 0, 8, 6, 9, 4],
[ 8, 4, 4, 1, 0, 1],
[ 4, 2, 5, 9, 4, 9],
[ 0, 6, 6, 3, 5, 4],
[50, 84, 96, 68, 78, 76],
[96, 51, 61, 65, 67, 57]])
nd1 = np.random.randint(10,20,size = (3,4))
nd2 = np.random.randint(40,60,size = (3,2))
np.concatenate((nd1,nd2), axis = 1)
array([[12, 17, 14, 12, 56, 44],
[13, 17, 17, 17, 51, 49],
[13, 16, 18, 15, 45, 59]])
- 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
nd3 = np.random.randint(0,10,size = (10,1))
nd3
array([[8],
[5],
[1],
[2],
[3],
[8],
[0],
[7],
[9],
[5]])
#hstack 水平级联
#horizontal :水平的
np.hstack(nd3)
#变成水平的之后,维度也变了
array([8, 5, 1, 2, 3, 8, 0, 7, 9, 5])
#vstack 垂直级联
#vertical :垂直的
nd4 = np.random.randint(-10,10,size = 10)
nd4
array([-10, -2, 3, 7, 1, 6, 6, -7, 0, -1])
np.vstack(nd4)
array([[-10],
[ -2],
[ 3],
[ 7],
[ 1],
[ 6],
[ 6],
[ -7],
[ 0],
[ -1]])
5. 切分
与级联类似,三个函数完成切分工作:
- np.split
- np.vsplit
- np.hsplit
nd = np.random.randint(0,100,size = (5,6))
nd
array([[44, 65, 84, 1, 83, 71],
[14, 71, 39, 21, 11, 27],
[27, 2, 89, 5, 13, 70],
[97, 63, 91, 45, 26, 71],
[86, 22, 3, 90, 56, 54]])
np.vsplit(nd,[1,4])
[array([[44, 65, 84, 1, 83, 71]]), array([[14, 71, 39, 21, 11, 27],
[27, 2, 89, 5, 13, 70],
[97, 63, 91, 45, 26, 71]]), array([[86, 22, 3, 90, 56, 54]])]
np.hsplit(nd,[1,3,8])
[array([[44],
[14],
[27],
[97],
[86]]), array([[65, 84],
[71, 39],
[ 2, 89],
[63, 91],
[22, 3]]), array([[ 1, 83, 71],
[21, 11, 27],
[ 5, 13, 70],
[45, 26, 71],
[90, 56, 54]]), array([], shape=(5, 0), dtype=int32)]
nd
array([[44, 65, 84, 1, 83, 71],
[14, 71, 39, 21, 11, 27],
[27, 2, 89, 5, 13, 70],
[97, 63, 91, 45, 26, 71],
[86, 22, 3, 90, 56, 54]])
np.split(nd,[2], axis = 1)
#axis = 0 默认的一种情况 行上面
[array([[44, 65],
[14, 71],
[27, 2],
[97, 63],
[86, 22]]), array([[84, 1, 83, 71],
[39, 21, 11, 27],
[89, 5, 13, 70],
[91, 45, 26, 71],
[ 3, 90, 56, 54]])]
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。
nd = np.random.randint(0,10,size = 6)
nd
array([9, 8, 4, 4, 1, 7])
nd[5] = 1000
nd
array([ 9, 8, 4, 4, 1, 1000])
可使用copy()函数创建副本
nd_copy = nd.copy()
nd_copy
array([ 9, 8, 4, 4, 1, 1000])
四、ndarray的聚合操作
1. 求和np.sum
np.power([2,3,4],3)
array([ 8, 27, 64], dtype=int32)
nd = np.random.randint(0,10,size = (3,4))
nd
array([[1, 1, 6, 7],
[2, 7, 0, 1],
[6, 6, 8, 0]])
np.power(nd, 2)
array([[ 1, 1, 36, 49],
[ 4, 49, 0, 1],
[36, 36, 64, 0]], dtype=int32)
nd = np.random.randint(0,10,size =(2,3))
nd
array([[4, 8, 6],
[0, 7, 0]])
nd.sum()/6
4.166666666666667
nd.sum(axis = 1)
array([18, 7])
nd.sum(axis = 0)
array([ 4, 15, 6])
nd.mean()
#求平均值
4.166666666666667
nd.mean(axis = 0)
array([2. , 7.5, 3. ])
nd
array([[4, 8, 6],
[0, 7, 0]])
nd.argmin()
3
nd1 = np.random.randint(12,34,size = (4,5))
nd1
array([[27, 25, 30, 30, 20],
[15, 30, 15, 27, 28],
[31, 13, 27, 12, 26],
[29, 22, 23, 15, 20]])
nd1.argmin()
#非常有用!!!!
13
nd1.argmax()
10
2. 最大最小值:np.max/ np.min
同理
nd1.max()
31
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
np.sum 和 np.nansum 的区别
nan not a number
操作文件
使用pandas打开文件president_heights.csv
获取文件中的数据
五、ndarray的矩阵操作
1. 基本矩阵操作
- 算术运算符:
- 加减乘除
nd = np.random.randint(0,10,size = (5,5))
nd
array([[6, 9, 2, 8, 5],
[5, 8, 3, 3, 6],
[3, 6, 3, 0, 5],
[4, 0, 6, 7, 2],
[6, 6, 8, 8, 5]])
nd + 3
#广播机制在里面
array([[ 9, 12, 5, 11, 8],
[ 8, 11, 6, 6, 9],
[ 6, 9, 6, 3, 8],
[ 7, 3, 9, 10, 5],
[ 9, 9, 11, 11, 8]])
nd/2
array([[3. , 4.5, 1. , 4. , 2.5],
[2.5, 4. , 1.5, 1.5, 3. ],
[1.5, 3. , 1.5, 0. , 2.5],
[2. , 0. , 3. , 3.5, 1. ],
[3. , 3. , 4. , 4. , 2.5]])
#系统还给咱们提供了一些方法
np.multiply(nd, 2)
#乘法
array([[12, 18, 4, 16, 10],
[10, 16, 6, 6, 12],
[ 6, 12, 6, 0, 10],
[ 8, 0, 12, 14, 4],
[12, 12, 16, 16, 10]])
np.subtract(nd,100)
#减法
array([[ -94, -91, -98, -92, -95],
[ -95, -92, -97, -97, -94],
[ -97, -94, -97, -100, -95],
[ -96, -100, -94, -93, -98],
[ -94, -94, -92, -92, -95]])
- 矩阵积np.dot()
矩阵乘法
nd1 = np.random.randint(0,10,size = (2,3))
nd2 = np.random.randint(0,10,size = (3,4))
np.dot(nd1,nd2)
array([[65, 54, 63],
[76, 66, 66]])
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
例2:
a = np.arange(3).reshape((3, 1))
b = np.arange(3)
求a+b
习题
a = np.ones((4, 1))
b = np.arange(4)
求a+b
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
#必须会默写至少两个排序
nd = np.random.randint(0,100,size = 10)
nd
array([40, 20, 21, 24, 88, 42, 30, 38, 35, 76])
def sort_nd(nd):
for i in range(nd.size):
for j in range(i, nd.size):
if nd[i] > nd[j]:
nd[i],nd[j] = nd[j],nd[i]
return nd
sort_nd(nd)
array([20, 21, 24, 30, 35, 38, 40, 42, 76, 88])
nd = np.random.randint(0,100,size = 10)
nd
array([12, 56, 48, 39, 64, 15, 58, 83, 10, 0])
def sort_nd2(nd):
for i in range(nd.size):
#argmin
#获取最小值的索引值
index_min = np.argmin(nd[i:]) + i
#当i = 0 index_min = 9
#当i = 1 index_min =
nd[i] ,nd[index_min]= nd[index_min],nd[i]
return nd
sort_nd2(nd)
array([ 0, 10, 12, 15, 39, 48, 56, 58, 64, 83])
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
nd = np.random.randint(0,100,size = 10)
nd
array([84, 73, 91, 38, 3, 56, 43, 70, 61, 72])
np.sort(nd)
array([ 3, 38, 43, 56, 61, 70, 72, 73, 84, 91])
nd
array([84, 73, 91, 38, 3, 56, 43, 70, 61, 72])
nd.sort()
nd
array([ 3, 38, 43, 56, 61, 70, 72, 73, 84, 91])
2. 部分排序
np.partition(a,k)
有的时候我们不是对全部数据感兴趣,我们可能只对最小或最大的一部分感兴趣。
- 当k为正时,我们想要得到最小的k个数
- 当k为负时,我们想要得到最大的k个数
nd = np.random.randint(0,1000,size = 500)
nd
array([679, 723, 152, 187, 847, 859, 843, 762, 239, 132, 183, 369, 168,
949, 533, 97, 480, 851, 309, 70, 140, 741, 383, 725, 478, 762,
553, 919, 935, 408, 295, 610, 601, 74, 986, 889, 600, 210, 945,
285, 209, 719, 111, 874, 347, 630, 978, 451, 500, 366, 773, 62,
506, 610, 619, 151, 667, 936, 234, 358, 846, 767, 865, 524, 126,
856, 832, 466, 428, 341, 474, 117, 891, 579, 287, 286, 947, 687,
368, 770, 838, 7, 246, 327, 513, 425, 794, 226, 144, 692, 423,
313, 457, 31, 900, 822, 781, 678, 548, 204, 687, 872, 134, 852,
264, 720, 894, 487, 780, 959, 633, 570, 54, 949, 336, 138, 319,
683, 115, 209, 56, 469, 326, 400, 362, 373, 726, 971, 948, 376,
575, 680, 122, 657, 961, 467, 586, 136, 763, 926, 533, 698, 960,
307, 609, 636, 649, 153, 308, 906, 520, 148, 465, 567, 231, 446,
456, 757, 388, 683, 946, 412, 671, 946, 959, 867, 673, 837, 518,
369, 494, 166, 808, 188, 253, 780, 511, 888, 332, 332, 8, 645,
779, 542, 998, 512, 287, 430, 835, 608, 759, 114, 740, 107, 552,
279, 885, 491, 346, 892, 739, 711, 908, 76, 233, 715, 915, 869,
673, 458, 21, 576, 297, 389, 35, 295, 25, 486, 664, 326, 260,
7, 87, 47, 242, 579, 889, 654, 465, 250, 364, 471, 758, 329,
579, 964, 774, 722, 710, 437, 763, 252, 551, 939, 765, 988, 186,
929, 767, 548, 583, 307, 775, 147, 936, 779, 959, 915, 673, 924,
456, 127, 472, 157, 287, 427, 449, 987, 174, 469, 148, 733, 846,
193, 725, 197, 988, 833, 498, 701, 696, 369, 915, 205, 81, 978,
218, 18, 984, 937, 169, 67, 617, 711, 177, 755, 691, 983, 360,
939, 313, 11, 54, 612, 626, 774, 442, 833, 547, 304, 967, 928,
85, 552, 231, 865, 227, 71, 997, 492, 484, 782, 498, 139, 361,
27, 925, 988, 842, 279, 185, 924, 932, 799, 972, 150, 107, 875,
949, 974, 445, 908, 733, 303, 909, 658, 941, 590, 14, 992, 800,
702, 409, 84, 62, 757, 865, 917, 711, 960, 448, 417, 961, 826,
215, 406, 208, 796, 12, 208, 86, 799, 533, 755, 806, 869, 245,
493, 128, 39, 572, 171, 951, 798, 101, 676, 715, 388, 707, 98,
35, 340, 397, 743, 166, 53, 568, 460, 545, 430, 349, 971, 370,
939, 138, 346, 96, 983, 393, 297, 615, 565, 805, 665, 435, 957,
991, 726, 489, 358, 86, 278, 124, 617, 643, 150, 583, 462, 658,
802, 848, 74, 807, 201, 354, 261, 408, 759, 361, 157, 829, 687,
963, 603, 617, 54, 306, 447, 952, 440, 972, 217, 808, 341, 586,
176, 852, 682, 770, 299, 108, 975, 440, 83, 807, 968, 131, 824,
428, 996, 556, 602, 159, 613, 711, 262, 342, 355, 191, 43, 666,
209, 766, 737, 829, 857, 263, 231, 992, 605, 479, 967, 168, 770,
885, 924, 986, 867, 130, 249])
np.partition(nd,20)
array([ 7, 8, 7, 27, 11, 21, 18, 25, 12, 14, 31, 43, 47,
54, 54, 35, 54, 53, 39, 35, 56, 76, 67, 74, 62, 74,
71, 62, 70, 81, 86, 84, 86, 85, 83, 87, 108, 148, 131,
205, 153, 176, 111, 115, 101, 98, 159, 183, 132, 138, 136, 171,
168, 197, 193, 151, 157, 148, 107, 150, 127, 166, 168, 169, 126,
201, 191, 174, 177, 147, 140, 117, 130, 150, 128, 185, 124, 166,
122, 97, 187, 152, 134, 188, 107, 157, 114, 204, 144, 138, 139,
208, 96, 186, 208, 297, 287, 227, 231, 226, 253, 304, 246, 279,
264, 278, 279, 286, 287, 261, 297, 234, 306, 295, 303, 218, 260,
217, 242, 209, 209, 285, 231, 299, 249, 210, 295, 250, 262, 215,
252, 209, 287, 307, 245, 263, 239, 231, 233, 307, 533, 533, 467,
493, 406, 376, 417, 388, 308, 448, 520, 373, 465, 400, 326, 446,
456, 469, 388, 409, 319, 412, 336, 445, 340, 397, 487, 460, 518,
369, 494, 545, 430, 349, 370, 361, 511, 346, 332, 332, 393, 498,
484, 542, 492, 512, 457, 430, 313, 423, 435, 489, 425, 513, 327,
358, 547, 491, 346, 442, 368, 462, 313, 474, 341, 360, 428, 466,
524, 458, 354, 408, 358, 389, 361, 506, 447, 486, 366, 326, 500,
451, 440, 347, 341, 440, 428, 369, 465, 408, 364, 471, 498, 329,
342, 355, 478, 469, 383, 437, 449, 309, 427, 480, 533, 472, 369,
479, 456, 362, 548, 583, 775, 723, 936, 779, 959, 915, 673, 924,
767, 929, 988, 765, 939, 551, 763, 987, 710, 722, 774, 733, 846,
964, 725, 579, 988, 833, 758, 701, 696, 654, 915, 889, 579, 978,
664, 576, 984, 937, 673, 869, 617, 711, 915, 755, 691, 983, 715,
939, 908, 711, 739, 612, 626, 774, 892, 833, 885, 552, 967, 928,
740, 552, 759, 865, 608, 835, 997, 998, 779, 782, 645, 888, 780,
808, 925, 988, 842, 837, 673, 924, 932, 799, 972, 867, 959, 875,
949, 974, 946, 908, 733, 671, 909, 658, 941, 590, 946, 992, 800,
702, 683, 757, 567, 757, 865, 917, 711, 960, 906, 649, 961, 826,
636, 609, 960, 796, 698, 926, 763, 799, 586, 755, 806, 869, 961,
657, 680, 575, 572, 948, 951, 798, 971, 676, 715, 726, 707, 683,
949, 570, 633, 743, 959, 780, 568, 894, 720, 852, 872, 971, 687,
939, 548, 678, 781, 983, 822, 900, 615, 565, 805, 665, 692, 957,
991, 726, 794, 838, 770, 687, 947, 617, 643, 579, 583, 891, 658,
802, 848, 832, 807, 856, 865, 767, 846, 759, 936, 667, 829, 687,
963, 603, 617, 619, 610, 773, 952, 978, 972, 630, 808, 874, 586,
719, 852, 682, 770, 945, 600, 975, 889, 986, 807, 968, 601, 824,
610, 996, 556, 602, 935, 613, 711, 919, 553, 762, 725, 741, 666,
851, 766, 737, 829, 857, 949, 762, 992, 605, 843, 967, 859, 770,
885, 924, 986, 867, 847, 679])
numpy基础用法学习的更多相关文章
- [学习笔记] Numpy基础 系统学习
[学习笔记] Numpy基础 上专业选修<数据分析程序设计>课程,老师串讲了Numpy基础,边听边用jupyter敲了下--理解+笔记. 老师讲的很全很系统,有些点没有记录,在PPT里就不 ...
- NumPy 基础用法
NumPy 是高性能科学计算和数据分析的基础包. 它是 pandas 等其他各种工具的基础. 主要功能: ndarray 一个多维数组结构, 高效且节省空间 无需循环对整组数据进行快速运算的数学函数 ...
- NumPy基础入门学习
对于习惯使用了MATLAB的用户而言,学习NumPy这个python工具包付出的成本应该是不大的. NumPy的基本的object是多维数组,是一个有同样类型的数字等构成的一张表格,能够通过元组进行索 ...
- 【Numpy】python机器学习包Numpy基础知识学习
一.安装:在之前的博客中已经写过:http://www.cnblogs.com/puyangsky/p/4763234.html 二.python数组切片知识: python中序列类有list.str ...
- JDBC 基础用法学习
JDBC概述 java 数据库链接,sun公司退出的 java 访问数据库的标准规范接口 是一种用于执行SQL语句的 java API 可以作为多种关系数据库提供统一接口 是一组 java 工具类和接 ...
- Python Numpy shape 基础用法(转自他人的博客,如涉及到侵权,请联系我)
Python Numpy shape 基础用法 shape函数是numpy.core.fromnumeric中的函数,它的功能是读取矩阵的长度,比如shape[0]就是读取矩阵第一维度的长度.它的输入 ...
- 【学习笔记】 第04章 NumPy基础:数组和矢量计算
前言 正式开始学习Numpy,参考用书是<用Python进行数据清洗>,计划本周五之前把本书读完,关键代码全部实现一遍 NumPy基础:数组和矢量计算 按照书中所示,要搞明白具体的性能差距 ...
- MarkDown学习——基础用法
目录 MarkDown开发版本MD2All基础用法 此处有代码<a id="top"></a>作为页内锚点 此处是用自动生成的目录 MarkDown是什么M ...
- python学习笔记(三):numpy基础
Counter函数可以对列表中数据进行统计每一个有多少种 most_common(10)可以提取前十位 from collections import Counter a = ['q','q','w' ...
随机推荐
- Email地址验证
<script>function validateForm(){ var x=document.forms["myForm"]["email"].v ...
- @loj - 3043@「ZJOI2019」线段树
目录 @description@ @solution@ @accepted code@ @details@ @description@ 九条可怜是一个喜欢数据结构的女孩子,在常见的数据结构中,可怜最喜 ...
- java中工厂模式
最近在项目中使用了工厂模式来重构下之前的代码,在这里做个小结. 工厂模式最主要的特点是每次新增一个产品的时候,都需要新增一个新的工厂,这样在对于新的产品做扩展的时候,减少对客户端代码的修改. 我在项目 ...
- numpy.stack和numpy.concatenate的区别
在使用numpy进行矩阵运算的时候踩到的坑,原因是不能正确区分numpy.concatenate和numpy.stack在功能上的差异. 先说numpy.concatenate,直接看文档: nump ...
- 《Java并发编程的艺术》第6/7/8章 Java并发容器与框架/13个原子操作/并发工具类
第6章 Java并发容器和框架 6.1 ConcurrentHashMap(线程安全的HashMap.锁分段技术) 6.1.1 为什么要使用ConcurrentHashMap 在并发编程中使用Has ...
- webpack简单笔记
本文简单记录学习webpack3.0的笔记,已备日后查阅.节省查阅文档时间 安装 可以使用npm安装 //全局安装 npm install -g webpack //安装到项目目录 npm insta ...
- 一对多分页的SQL到底应该怎么写?
1. 前言 MySQL一对多的数据分页是非常常见的需求,比如我们要查询商品和商品的图片信息.但是很多人会在这里遇到分页的误区,得到不正确的结果.今天就来分析并解决这个问题. 2. 问题分析 我们先创建 ...
- jvm基础知识1
堆放实例对象,栈放实例对象的引用,方法区存储创建类的信息 上面堆和垃圾回收的关系,垃圾回收回收的是堆内存的数据,s0和s1区域, 例如现在我们要清除s0中的堆对象,将s0中正在运行的对象从s0区域移动 ...
- Python 简明教程 --- 17,Python 模块与包
微信公众号:码农充电站pro 个人主页:https://codeshellme.github.io 正确的判断来源于经验,然而经验来源于错误的判断. -- Fred Brooks 目录 我们已经知道函 ...
- 深入理解RocketMQ(四)--消息存储
一.MQ存储分类 MQ存储主要分为以下三类: 文件系统:RocketMQ/Kafka/RabbitMQ 关系型数据库DB:ActiveMQ(默认采用的KahaDB做消息存储)可选用JDBC的方式来做消 ...