numpy学习笔记

import numpy  as np

N维数组对象ndarray

np.array()生成一个ndarray数组

轴(axis)保存数据维度，秩(rank)轴的数量

ndarray对象的属性：

避免使用非同质的ndarray对象。

np.array(list/tuple,dtype=np.float32)

In [24]: np.arange(10)
Out[24]: array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])

In [25]: np.ones((3,4))
Out[25]:
array([[ 1., 1., 1., 1.],
[ 1., 1., 1., 1.],
[ 1., 1., 1., 1.]])

In [26]: np.ones((3,4),dtype=np.int32)
Out[26]:
array([[1, 1, 1, 1],
[1, 1, 1, 1],
[1, 1, 1, 1]], dtype=int32)

In [28]: np.zeros((2,3))
Out[28]:
array([[ 0., 0., 0.],
[ 0., 0., 0.]])

In [29]: np.eye(5)
Out[29]:
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.]])

In [40]: a=np.linspace(1,25,5)

In [41]: a
Out[41]: array([  1.,   7.,  13.,  19.,  25.])

In [42]: b=np.linspace(1,25,5,endpoint=False)

In [43]: b
Out[43]: array([  1. ,   5.8,  10.6,  15.4,  20.2])

In [45]: np.concatenate((a,b))
Out[45]: array([  1. ,   7. ,  13. ,  19. ,  25. ,   1. ,   5.8,  10.6,  15.4,  20.2])

ndarray数组维度的变换

In [47]: a=np.ones((2,3,4),dtype=np.int32)

In [48]: a
Out[48]: 
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]]], dtype=int32)

In [49]: a.reshape(4,6)
Out[49]: 
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]], dtype=int32)

In [50]: a
Out[50]: 
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]]], dtype=int32)

In [51]: a.resize((4,6))

In [52]: a
Out[52]: 
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]], dtype=int32)

In [53]: a.flatten()
Out[53]: array([1, 1, 1, ..., 1, 1, 1], dtype=int32)

In [54]: a
Out[54]: 
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]], dtype=int32)

reshape()是另返回一个新的array，原array不变；resize()是改变原数组

数组类型的改变：astype()

ndarray数组向列表的转换：tolist()

In [2]: a=np.ones((2,3,4),dtype=np.int32) 

In [3]: a
Out[3]:  
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]]], dtype=int32)

In [4]: b=a.astype(np.float)
In [5]: b
Out[5]:  
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.]]])

数据的csv文件存取

　　np.savetxt(frame,array,fmt='%0.18e',delimiter=None)

In [22]: a=np.arange(15).reshape((3,5))

In [24]: np.savetxt('a.csv',a,fmt='%d',delimiter=',')

　　np.loadtxt(frame,dtype=np.float,delimiter=None,unpack=False)

In [25]: b=np.loadtxt('a.csv',delimiter=',')

In [26]: b
Out[26]: 
array([[  0.,   1.,   2.,   3.,   4.],
       [  5.,   6.,   7.,   8.,   9.],
       [ 10.,  11.,  12.,  13.,  14.]])

In [27]: b=np.loadtxt('a.csv',dtype=np.int,delimiter=',')

In [28]: b
Out[28]: 
array([[ 0,  1,  2,  3,  4],
       [ 5,  6,  7,  8,  9],
       [10, 11, 12, 13, 14]])

 以上两个函数只能有效存取一维和二维数组，

对于多维数据的存取，可使用tofile()和fromfile()，这两个方法需要知道存入文件时数组的维度和元素类型。

In [29]: a=np.arange(100).reshape(5,2,10)

In [30]: a
Out[30]: 
array([[[ 0,  1,  2, ...,  7,  8,  9],
        [10, 11, 12, ..., 17, 18, 19]],

       [[20, 21, 22, ..., 27, 28, 29],
        [30, 31, 32, ..., 37, 38, 39]],

       [[40, 41, 42, ..., 47, 48, 49],
        [50, 51, 52, ..., 57, 58, 59]],

       [[60, 61, 62, ..., 67, 68, 69],
        [70, 71, 72, ..., 77, 78, 79]],

       [[80, 81, 82, ..., 87, 88, 89],
        [90, 91, 92, ..., 97, 98, 99]]])

In [32]: a.tofile('b.dat',sep=',',format="%d")

In [37]: c=np.fromfile('b.dat',dtype=np.int,sep=',')

In [38]: c
Out[38]: array([ 0,  1,  2, ..., 97, 98, 99])

In [39]: c=np.fromfile('b.dat',dtype=np.int,sep=',').reshape(5,2,10)

In [40]: c
Out[40]: 
array([[[ 0,  1,  2, ...,  7,  8,  9],
        [10, 11, 12, ..., 17, 18, 19]],

       [[20, 21, 22, ..., 27, 28, 29],
        [30, 31, 32, ..., 37, 38, 39]],

       [[40, 41, 42, ..., 47, 48, 49],
        [50, 51, 52, ..., 57, 58, 59]],

       [[60, 61, 62, ..., 67, 68, 69],
        [70, 71, 72, ..., 77, 78, 79]],

       [[80, 81, 82, ..., 87, 88, 89],
        [90, 91, 92, ..., 97, 98, 99]]])

此外，还可以使用np.save(frame,array)、np.savez(frame,array)和np.load(frame)。这种方法很便捷，以npy为扩展名（压缩扩展名为npz），自动保存了维度信息。

In [42]: a=np.arange(20).reshape(2,2,5)

In [44]: np.save('a.npy',a)

In [47]: b=np.load('a.npy')

In [48]: b
Out[48]: 
array([[[ 0,  1,  2,  3,  4],
        [ 5,  6,  7,  8,  9]],

       [[10, 11, 12, 13, 14],
        [15, 16, 17, 18, 19]]])

 np.random的随机数函数

In [2]: a=np.random.rand(2,3,4)

In [3]: a
Out[3]: 
array([[[ 0.3774548 ,  0.38378479,  0.23116153,  0.62468074],
        [ 0.54844696,  0.07423047,  0.6164151 ,  0.20719271],
        [ 0.70499044,  0.49245108,  0.41352731,  0.21213278]],

       [[ 0.6749494 ,  0.10447893,  0.88275619,  0.85359191],
        [ 0.75249162,  0.18598287,  0.40681266,  0.55572018],
        [ 0.52925702,  0.05278294,  0.45759326,  0.20160628]]])

In [4]: b=np.random.randint(10,20,(2,3,4))

In [5]: b
Out[5]: 
array([[[12, 16, 16, 13],
        [14, 10, 15, 16],
        [18, 18, 13, 16]],

       [[15, 19, 15, 13],
        [17, 17, 19, 17],
        [13, 19, 10, 12]]])

In [8]: u=np.random.uniform(0,10,(3,4))

In [9]: u
Out[9]: 
array([[ 5.43561499,  3.36781091,  5.66832985,  8.51846059],
       [ 5.85174133,  3.46196561,  9.50727415,  0.95776365],
       [ 2.63730522,  8.21679612,  4.88207727,  1.80323693]])

In [10]: n=np.random.normal(10,5,(3,4))

In [11]: n
Out[11]: 
array([[ 10.41948308,   3.57014388,   7.1968483 ,   1.41896386],
       [  6.86452724,   4.73895058,  13.04939017,  10.82616448],
       [  7.92904101,  10.10575646,  11.47346598,  12.61651639]])