numpy-learn

发布时间:2023年12月19日

创建数组

import numpy as np
import pandas as pd
import math

value = float('nan')

# 使用 math.isnan()
if math.isnan(value):
    print("Value is NaN")

# 使用 numpy.isnan()
if np.isnan(value):
    print("Value is NaN")

np.array([1, 2, 3, 4, 5])
np.linspace(10, 100, 10)

Value is NaN
Value is NaN





array([ 10.,  20.,  30.,  40.,  50.,  60.,  70.,  80.,  90., 100.])
sex = pd.Series(['Male','Male','Female'])
np.array(sex)
array(['Male', 'Male', 'Female'], dtype=object)

linspace

start:起始数字

end:结束

Num:要生成的样本数,默认为50。

np.linspace(10,100,10)

Arange: step:数值步长

np.arange(5,10,2)
array([5, 7, 9])

Uniform: 在上下限之间的均匀分布中生成随机样本

np.random.uniform(5,10,size = 4)
array([5.35806766, 5.25970119, 5.53573947, 7.04110989])
np.random.uniform(size = 5)
array([0.53490428, 0.07574269, 0.2994071 , 0.10866207, 0.78867775])
np.random.uniform(size = (2,3))
array([[0.23929405, 0.26832237, 0.70498685],
       [0.71195525, 0.50311203, 0.99720624]])

Random.randint:在一个范围内生成n个随机整数样本

np.random.randint(5,10,10)
array([9, 8, 6, 6, 9, 5, 9, 5, 6, 5])

Random.random:生成n个随机浮点数样本

np.random.random(3)
array([0.70025574, 0.50566006, 0.80192119])

Logspace:在对数尺度上生成间隔均匀的数字
Start:序列的起始值。

End:序列的最后一个值。

endpoint:如果为True,最后一个样本将包含在序列中。

base:底数。默认是10

np.logspace(0,10,5,base=2)
array([1.00000000e+00, 5.65685425e+00, 3.20000000e+01, 1.81019336e+02,
       1.02400000e+03])

zeroes: np.zeroes会创建一个全部为0的数组,

shape:阵列的形状。

Dtype:生成数组所需的数据类型。’ int ‘或默认’ float ’

np.zeros((2,3),dtype="int")
array([[0, 0, 0],
       [0, 0, 0]])
np.zeros(5)
array([0., 0., 0., 0., 0.])

ones: np.ones函数创建一个全部为1的数组

np.ones((2,3))
array([[1., 1., 1.],
       [1., 1., 1.]])
full:创建一个单独值的n维数组
fill_value:填充值
np.full((2,4),fill_value=4)
array([[4, 4, 4, 4],
       [4, 4, 4, 4]])

Identity:创建具有指定维度的单位矩阵

np.identity(4)
array([[1., 0., 0., 0.],
       [0., 1., 0., 0.],
       [0., 0., 1., 0.],
       [0., 0., 0., 1.]])

数组操作

max,min:返回数组中的最小值
axis:用于操作的轴
out:用于存储输出的数组

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

unique:返回一个所有唯一元素排序的数组
return_index:如果为True,返回数组的索引。

return_inverse:如果为True,返回唯一数组的下标。

return_counts:如果为True,返回数组中每个唯一元素出现的次数。

axis:要操作的轴。默认情况下,数组被认为是扁平的
np.unique(arr,return_counts=True)
(array([1, 2, 3, 4, 5, 6]), array([2, 2, 2, 1, 1, 2]))

mean: 返回数组的平均数

np.mean(arr)
3.3

medain: 返回中位数

arr = np.array([[1,2,3],[5,8,4]])
np.median(arr)
3.5

digitize: 返回输入数组中每个值所属的容器的索引。bin:容器的数组,right:表示该间隔是否包括右边或左边的bin

a = np.array([-0.9, 0.5, 0.9, 1, 1.2, 1.4, 3.6, 4.7, 5.3])
bin = np.array([0,1,2,3])
np.digitize(a,bin)
array([0, 1, 1, 2, 2, 2, 4, 4, 4])
reshape:它返回一个数组,其中包含具有新形状的相同数据
A = np.random.randint(15,size=(4,3))
A
array([[ 0,  3,  3],
       [12, 12,  7],
       [ 7,  8, 11],
       [10, 14,  9]])
A.reshape(3,4)
array([[ 0,  3,  3, 12],
       [12,  7,  7,  8],
       [11, 10, 14,  9]])
A.reshape(-1)
array([ 0,  3,  3, 12, 12,  7,  7,  8, 11, 10, 14,  9])

expand_dims:它用于扩展数组的维度

arr = np.array([ 8, 14, 1, 8, 11, 4, 9, 4, 1, 13, 13, 11])
np.expand_dims(arr,axis=0)
array([[ 8, 14,  1,  8, 11,  4,  9,  4,  1, 13, 13, 11]])
np.expand_dims(arr,axis=1)

array([[ 8],
       [14],
       [ 1],
       [ 8],
       [11],
       [ 4],
       [ 9],
       [ 4],
       [ 1],
       [13],
       [13],
       [11]])
arr = np.array([1, 2, 3])  # 输入数组
result = np.expand_dims(arr, axis=1)
result
array([[1],
       [2],
       [3]])

squeeze:通过移除一个单一维度来降低数组的维度

arr = np.array([[ 8],[14],[ 1],[ 8],[11],[ 4],[ 9],[ 4],[ 1],[13],[13],[11]])
np.squeeze(arr)
array([ 8, 14,  1,  8, 11,  4,  9,  4,  1, 13, 13, 11])

count_nonzero: 计算所有非零元素并返回它们的计数

a = np.array([0,0,1,1,1,0])
np.count_nonzero(a)
3

argwhere: 查找并返回非零元素的所有下标

a = np.array([0,0,1,1,1,0])
np.argwhere(a)
array([[2],
       [3],
       [4]])

argmax & argmin:argmax返回数组中Max元素的索引。它可以用于多类图像分类问题中获得高概率预测标签的指标,argmin将返回数组中min元素的索引

arr = np.array([[0.12,0.64,0.19,0.05]])
np.argmax(arr)
1
np.argmin(arr)
3

sort: 对数组排序,kind:要使用的排序算法。{‘quicksort’, ‘mergesort’, ‘heapsort’, ‘stable’}

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

abs:返回数组中元素的绝对值

A=np.array([[1,-3,4],[-2,-4,3]])
np.abs(A)
array([[1, 3, 4],
       [2, 4, 3]])

round:将浮点值四舍五入到指定数目的小数点,decimals:要保留的小数点的个数

a=np.random.random(size=(3,4))
a

array([[0.16884867, 0.57913567, 0.16815851, 0.20774758],
       [0.54647561, 0.9234027 , 0.83512324, 0.06706385],
       [0.70642496, 0.02558962, 0.08868171, 0.52844368]])

clip: 它可以将数组的裁剪值保持在一个范围内

arr = np.array([0,1,-3,-4,5,6,7,2,3])
arr.clip(0,5)
array([0, 1, 0, 0, 5, 5, 5, 2, 3])

替换数组中的值

where:返回满足条件的数组元素,condition:匹配的条件。如果true则返回x,否则y。

a = np.arange(12).reshape(4,3)
a
array([[ 0,  1,  2],
       [ 3,  4,  5],
       [ 6,  7,  8],
       [ 9, 10, 11]])
np.where(a>5)
(array([2, 2, 2, 3, 3, 3]), array([0, 1, 2, 0, 1, 2]))
a[np.where(a>5)]
array([ 6,  7,  8,  9, 10, 11])

可以用来替换pandas df中的元素

put:用给定的值替换数组中指定的元素,
a:数组

Ind:需要替换的索引。

V:替换值

arr = np.array([1,2,3,4,5,6])
arr
array([1, 2, 3, 4, 5, 6])
np.put(arr,[1,2],[10,9])
arr
array([ 1, 10,  9,  4,  5,  6])

copyto:将一个数组的内容复制到另一个数组中

arr1 = np.array([1,2,3])
arr2 = np.array([4,5,6])
arr1
array([1, 2, 3])
arr2
array([4, 5, 6])
np.copyto(dst=arr1,src=arr2)
arr2
array([4, 5, 6])
arr1
array([4, 5, 6])

集合操作

查找公共元素: intersect1d函数以排序的方式返回两个数组中所有唯一的值
Assume_unique:如果为真值,则假设输入数组都是唯一的。

Return_indices:如果为真,则返回公共元素的索引
ar1 = np.array([1,2,3,4,5,6])
ar2 = np.array([3,4,5,8,9,1])
np.intersect1d(ar1,ar2)
array([1, 3, 4, 5])

查找不同元素: np.setdiff1d函数返回arr1中在arr2中不存在的所有唯一元素

a = np.array([1, 7, 3, 2, 4, 1])
b = np.array([9, 2, 5, 6, 7, 8])
np.setdiff1d(a,b)
array([1, 3, 4])

从两个数组中提取唯一元素:Setxor1d 将按顺序返回两个数组中所有唯一的值

a = np.array([1, 2, 3, 4, 6])
b = np.array([1, 4, 9, 4, 36])

np.setxor1d(a,b)
array([ 2,  3,  6,  9, 36])

合并:Union1d函数将两个数组合并为一个

a = np.array([1, 2, 3, 4, 5])
b = np.array([1, 3, 5, 4, 36])
np.union1d(a,b)
array([ 1,  2,  3,  4,  5, 36])

数组分割:

Hsplit函数将数据水平分割为n个相等的部分

A = np.array([[3,4,5,2],[6,7,2,6]])
np.hsplit(A,2)
[array([[3, 4],
        [6, 7]]),
 array([[5, 2],
        [2, 6]])]
np.hsplit(A,4)
[array([[3],
        [6]]),
 array([[4],
        [7]]),
 array([[5],
        [2]]),
 array([[2],
        [6]])]

垂直分割:Vsplit将数据垂直分割为n个相等的部分

A = np.array([[3,4,5,2],[6,7,2,6]])
np.vsplit(A,2)
[array([[3, 4, 5, 2]]), array([[6, 7, 2, 6]])]

数组叠加

水平叠加:hstack 将在另一个数组的末尾追加一个数组

a = np.array([1,2,3,4,5])
b = np.array([1,4,9,16,25])
np.hstack((a,b))
array([ 1,  2,  3,  4,  5,  1,  4,  9, 16, 25])

垂直叠加:vstack将一个数组堆叠在另一个数组上

np.vstack((a,b))
array([[ 1,  2,  3,  4,  5],
       [ 1,  4,  9, 16, 25]])

allclose:如果两个数组的形状相同,则Allclose函数根据公差值查找两个数组是否相等或近似相等

a = np.array([0.25,0.4,0.6,0.32])
b = np.array([0.26,0.3,0.7,0.32])
tolerance = 0.1 ## Total Difference 
np.allclose(a,b,tolerance)
False
tolerance = 0.5
np.allclose(a,b,tolerance)
True

equal:它比较两个数组的每个元素,如果元素匹配就返回True

np.equal(arr1,arr2)
array([ True,  True,  True])

重复的数组元素

repeat:它用于重复数组中的元素n次,

A:重复的元素

Repeats:重复的次数

np.repeat('2017',3)
array(['2017', '2017', '2017'], dtype='<U4')
fruits=pd.DataFrame([
['Mango',40],
['Apple',90],
['Banana',130]
],columns=['Product','ContainerSales'])

fruits
ProductContainerSales
0Mango40
1Apple90
2Banana130

[0] 表示索引操作,用于访问数组或矩阵的维度信息

fruits['year'] = np.repeat(2020,fruits.shape[0])
fruits
ProductContainerSalesyear
0Mango402020
1Apple902020
2Banana1302020

tile: 通过重复A,rep次来构造一个数组

np.tile("Ram",5)
array(['Ram', 'Ram', 'Ram', 'Ram', 'Ram'], dtype='<U3')
np.tile(3,(2,3))
array([[3, 3, 3],
       [3, 3, 3]])

爱因斯坦求和

a = np.arange(1,10).reshape(3,3)
b = np.arange(21,30).reshape(3,3)
np.einsum('ii->i',a)
array([1, 5, 9])
np.einsum('ji',a)
array([[1, 4, 7],
       [2, 5, 8],
       [3, 6, 9]])
np.einsum('ij,jk',a,b)
array([[150, 156, 162],
       [366, 381, 396],
       [582, 606, 630]])
np.einsum('ii',a)
15

统计分析

A = np.array([[3, 4, 5, 2],
[6, 7, 2, 6]])
np.histogram(A)
(array([2, 0, 1, 0, 1, 0, 1, 0, 2, 1]),
 array([2. , 2.5, 3. , 3.5, 4. , 4.5, 5. , 5.5, 6. , 6.5, 7. ]))

百分位数

沿指定轴计算数据的Q-T-T百分位数
a:输入
q:要计算的百分位。

overwrite_input:如果为true,则允许输入数组修改中间计算以节省内存
a = np.array([[2, 4, 6], [4, 8, 12]])
np.percentile(a, 50)
5.0
np.percentile(a, 10)
3.0
np.percentile(a,5)
2.5

标准偏差和方差
std和var是NumPy的两个函数,用于计算沿轴的标准偏差和方差

a = np.array([[2, 4, 6], [4, 8, 12]])
np.std(a,axis=1)
array([1.63299316, 3.26598632])
np.std(a,axis=0)
array([1., 2., 3.])
np.var(a,axis=1)
array([ 2.66666667, 10.66666667])
np.var(a,axis=0)
array([1., 4., 9.])

数组打印

np.set_printoptions(precision=2)
a = np.array([12.23456,32.34535])
a
array([12.23, 32.35])

设置打印数组最大值

np.set_printoptions(threshold=np.inf)

增加一行中元素的数量

np.set_printoptions(linewidth=100)

保存和加载数据

arr = np.linspace(10,100,500).reshape(25,20) 
np.savetxt('array.txt',arr)

加载

np.loadtxt('array.txt')
array([[ 10.  ,  10.18,  10.36,  10.54,  10.72,  10.9 ,  11.08,  11.26,  11.44,  11.62,  11.8 ,
         11.98,  12.16,  12.34,  12.53,  12.71,  12.89,  13.07,  13.25,  13.43],
       [ 13.61,  13.79,  13.97,  14.15,  14.33,  14.51,  14.69,  14.87,  15.05,  15.23,  15.41,
         15.59,  15.77,  15.95,  16.13,  16.31,  16.49,  16.67,  16.85,  17.03],
       [ 17.21,  17.39,  17.58,  17.76,  17.94,  18.12,  18.3 ,  18.48,  18.66,  18.84,  19.02,
         19.2 ,  19.38,  19.56,  19.74,  19.92,  20.1 ,  20.28,  20.46,  20.64],
       [ 20.82,  21.  ,  21.18,  21.36,  21.54,  21.72,  21.9 ,  22.08,  22.26,  22.44,  22.63,
         22.81,  22.99,  23.17,  23.35,  23.53,  23.71,  23.89,  24.07,  24.25],
       [ 24.43,  24.61,  24.79,  24.97,  25.15,  25.33,  25.51,  25.69,  25.87,  26.05,  26.23,
         26.41,  26.59,  26.77,  26.95,  27.13,  27.31,  27.49,  27.68,  27.86],
       [ 28.04,  28.22,  28.4 ,  28.58,  28.76,  28.94,  29.12,  29.3 ,  29.48,  29.66,  29.84,
         30.02,  30.2 ,  30.38,  30.56,  30.74,  30.92,  31.1 ,  31.28,  31.46],
       [ 31.64,  31.82,  32.  ,  32.18,  32.36,  32.55,  32.73,  32.91,  33.09,  33.27,  33.45,
         33.63,  33.81,  33.99,  34.17,  34.35,  34.53,  34.71,  34.89,  35.07],
       [ 35.25,  35.43,  35.61,  35.79,  35.97,  36.15,  36.33,  36.51,  36.69,  36.87,  37.05,
         37.23,  37.41,  37.6 ,  37.78,  37.96,  38.14,  38.32,  38.5 ,  38.68],
       [ 38.86,  39.04,  39.22,  39.4 ,  39.58,  39.76,  39.94,  40.12,  40.3 ,  40.48,  40.66,
         40.84,  41.02,  41.2 ,  41.38,  41.56,  41.74,  41.92,  42.1 ,  42.28],
       [ 42.46,  42.65,  42.83,  43.01,  43.19,  43.37,  43.55,  43.73,  43.91,  44.09,  44.27,
         44.45,  44.63,  44.81,  44.99,  45.17,  45.35,  45.53,  45.71,  45.89],
       [ 46.07,  46.25,  46.43,  46.61,  46.79,  46.97,  47.15,  47.33,  47.52,  47.7 ,  47.88,
         48.06,  48.24,  48.42,  48.6 ,  48.78,  48.96,  49.14,  49.32,  49.5 ],
       [ 49.68,  49.86,  50.04,  50.22,  50.4 ,  50.58,  50.76,  50.94,  51.12,  51.3 ,  51.48,
         51.66,  51.84,  52.02,  52.2 ,  52.38,  52.57,  52.75,  52.93,  53.11],
       [ 53.29,  53.47,  53.65,  53.83,  54.01,  54.19,  54.37,  54.55,  54.73,  54.91,  55.09,
         55.27,  55.45,  55.63,  55.81,  55.99,  56.17,  56.35,  56.53,  56.71],
       [ 56.89,  57.07,  57.25,  57.43,  57.62,  57.8 ,  57.98,  58.16,  58.34,  58.52,  58.7 ,
         58.88,  59.06,  59.24,  59.42,  59.6 ,  59.78,  59.96,  60.14,  60.32],
       [ 60.5 ,  60.68,  60.86,  61.04,  61.22,  61.4 ,  61.58,  61.76,  61.94,  62.12,  62.3 ,
         62.48,  62.67,  62.85,  63.03,  63.21,  63.39,  63.57,  63.75,  63.93],
       [ 64.11,  64.29,  64.47,  64.65,  64.83,  65.01,  65.19,  65.37,  65.55,  65.73,  65.91,
         66.09,  66.27,  66.45,  66.63,  66.81,  66.99,  67.17,  67.35,  67.54],
       [ 67.72,  67.9 ,  68.08,  68.26,  68.44,  68.62,  68.8 ,  68.98,  69.16,  69.34,  69.52,
         69.7 ,  69.88,  70.06,  70.24,  70.42,  70.6 ,  70.78,  70.96,  71.14],
       [ 71.32,  71.5 ,  71.68,  71.86,  72.04,  72.22,  72.4 ,  72.59,  72.77,  72.95,  73.13,
         73.31,  73.49,  73.67,  73.85,  74.03,  74.21,  74.39,  74.57,  74.75],
       [ 74.93,  75.11,  75.29,  75.47,  75.65,  75.83,  76.01,  76.19,  76.37,  76.55,  76.73,
         76.91,  77.09,  77.27,  77.45,  77.64,  77.82,  78.  ,  78.18,  78.36],
       [ 78.54,  78.72,  78.9 ,  79.08,  79.26,  79.44,  79.62,  79.8 ,  79.98,  80.16,  80.34,
         80.52,  80.7 ,  80.88,  81.06,  81.24,  81.42,  81.6 ,  81.78,  81.96],
       [ 82.14,  82.32,  82.51,  82.69,  82.87,  83.05,  83.23,  83.41,  83.59,  83.77,  83.95,
         84.13,  84.31,  84.49,  84.67,  84.85,  85.03,  85.21,  85.39,  85.57],
       [ 85.75,  85.93,  86.11,  86.29,  86.47,  86.65,  86.83,  87.01,  87.19,  87.37,  87.56,
         87.74,  87.92,  88.1 ,  88.28,  88.46,  88.64,  88.82,  89.  ,  89.18],
       [ 89.36,  89.54,  89.72,  89.9 ,  90.08,  90.26,  90.44,  90.62,  90.8 ,  90.98,  91.16,
         91.34,  91.52,  91.7 ,  91.88,  92.06,  92.24,  92.42,  92.61,  92.79],
       [ 92.97,  93.15,  93.33,  93.51,  93.69,  93.87,  94.05,  94.23,  94.41,  94.59,  94.77,
         94.95,  95.13,  95.31,  95.49,  95.67,  95.85,  96.03,  96.21,  96.39],
       [ 96.57,  96.75,  96.93,  97.11,  97.29,  97.47,  97.66,  97.84,  98.02,  98.2 ,  98.38,
         98.56,  98.74,  98.92,  99.1 ,  99.28,  99.46,  99.64,  99.82, 100.  ]])

refer

文章来源:https://blog.csdn.net/weixin_38805083/article/details/135071753
本文来自互联网用户投稿,该文观点仅代表作者本人,不代表本站立场。本站仅提供信息存储空间服务,不拥有所有权,不承担相关法律责任。