什么是numpy?
一个在Python中做科学计算的基础库，重在数值计算，也是大部分PYTHON科学计算库的基础库，多用于在大型、多维数组上执行数值运算.
numpy是用来处理数值型数据的
import random import numpy as np # 使用numpy生成数组,生成numpy.ndarray的类型,这种类型数值之间没有逗号 # np.array(p_object,dtype=None) # p_object列表 dtype指定数据类型 t1 = np.array([1, 2, 3]) print(t1, type(t1)) # 结果[1 2 3] <class 'numpy.ndarray'> t2 = np.array(range(10)) print(t2) # 结果 [0 1 2 3 4 5 6 7 8 9] # arrange使用方法 arange([start,] stop [,step], dtype=None) # dtype指定是当前数组的类型 t3 = np.arange(10) # 帮助我们生成一堆数组 print(t3) # 结果 [0 1 2 3 4 5 6 7 8 9] # dtype 查看当前数组的类型 print(t3.dtype) # 结果 int64 64当前电脑的位数,存一个数值需要64位,位数越少占用内存越少 t4 = np.array([1, 3, 5], dtype=bool) print(t4) # 结果 [0 1 2 3 4 5 6 7 8 9] [ True True True] # astype 复制调原有数组并更改数据类型 t5 = t4.astype('int64') print(t5) # 结果 [1 1 1] # numpy中的小数 t6 = np.array([random.random() for i in range(10)]) print(t6) # 结果 [0.66398033 0.59250261 0.46354278 0.66351646 0.07548666 0.06377063 0.01887494 0.63553692 0.85932655 0.0017619 ] # 保留两位小数 t7 = t6.round(2) print(t7) # 结果 [0.29 0.92 0.94 0.62 0.26 0.43 0.25 0.33 0.1 0.4 ]
数组的形状
# 1.查看数组的形状 shape In [5]: t1.shape Out[5]: (12,) #2.修改数组的形状 reshape,不修改原数组 In [13]: t2 = np.arange(12) In [14]: t2 Out[14]: array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11]) In [15]: t3 = t2.reshape(3,4) In [16]: t3 Out[16]: array([[ 0, 1, 2, 3], [ 4, 5, 6, 7], [ 8, 9, 10, 11]])
# 查看数组的维度
array.ndim
多维数组
#将一维数组变成3纬数组 In [19]: t5 = np.arange(24) In [20]: t5 Out[20]: 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]) In [21]: t5.reshape(2,3,4) # 2代表2块数据,3代表每块3行,4代表每块4列 Out[21]: 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]]]) #变成一维数组 t6.reshape((24,)) # 而不是t6.reshape((24,1)) In [23]: t6 Out[23]: 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]]]) In [24]: t6.flatten() # 直接将t6变成一维的数组 Out[24]: 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])
数组计算
#任何数和数组计算,都相当于和数组中的每个数计算,这是广播现象 In [26]: t7 Out[26]: 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]) In [27]: t7+2 Out[27]: array([ 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25]) # 两个数组相加,对应位置上的数相加 In [28]: t1 = np.arange(100,124) In [29]: t1 Out[29]: array([100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123]) In [30]: t7 Out[30]: 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]) In [31]: t7+t1 Out[31]: array([100, 102, 104, 106, 108, 110, 112, 114, 116, 118, 120, 122, 124, 126, 128, 130, 132, 134, 136, 138, 140, 142, 144, 146]) # 两个维度不相同的数计算(前提列数相同) In [32]: t1 =np.arange(1,5) In [33]: t1 Out[33]: array([1, 2, 3, 4]) In [34]: t2 = np.arange(1,4) In [35]: t2 Out[35]: array([1, 2, 3]) In [36]: t1+t2 --------------------------------------------------------------------------- ValueError Traceback (most recent call last) <ipython-input-36-5260db44d29f> in <module> ----> 1 t1+t2 ValueError: operands could not be broadcast together with shapes (4,) (3,) #正确的,多维的那个数组中的每行数据都会和一维的那个数据计算 In [39]: t2 = np.arange(0,8).reshape(2,4) In [40]: t2 Out[40]: array([[0, 1, 2, 3], [4, 5, 6, 7]]) In [41]: t1 Out[41]: array([1, 2, 3, 4]) In [42]: t2+t1 Out[42]: array([[ 1, 3, 5, 7], [ 5, 7, 9, 11]])
numpy的广播原则:
如果两个数组的后缘维度(从末尾开始算起的维度)的轴长度相符或其中一方的长度为1，则认为它们是广播兼容的。广播会在缺失维度和(或)轴长度为1的维度上进行。
这是官方的话,我们该怎么理解呢?举个例子
shape为(3,)每行中有3列和shape为(2,3)每行中也是有3列的,这样的是可以计算的,也就是说两个数组中的列数只要相同就可以计算
数组的转置
transpose 将行列对调
T 也有这个功能
In [46]: t2 = np.arange(24).reshape(4,6) In [47]: t2 Out[47]: 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]]) In [48]: t2.transpose() Out[48]: array([[ 0, 6, 12, 18], [ 1, 7, 13, 19], [ 2, 8, 14, 20], [ 3, 9, 15, 21], [ 4, 10, 16, 22], [ 5, 11, 17, 23]]) In [49]:
浙公网安备 33010602011771号