Python之Numpy学习（1）
astype(dtype):显示的转换数组的数据类型，该方法总会生成一个新数组
In [6]: arr = np.arange(10,dtype=np.float)
In [7]: arr
Out[7]: array([0., 1., 2., 3., 4., 5., 6., 7., 8., 9.])
In [8]: arr = arr.astype(np.int)
In [9]: arr
Out[9]: array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])对于数组元素是字符串也同样适用：
In [10]: s = np.array(['1.24','3.56','24'])
In [11]: s = s.astype(float)
In [12]: s
Out[12]: array([ 1.24, 3.56, 24. ])对于numpy的标量算术运算会把参数传递给数组的每一个元素
In [18]: arr1 = np.arange(12).reshape(3,4)
In [19]: arr2 = np.arange(12,24).reshape(3,4)
In [20]: arr1
Out[20]:
array([[ 0, 1, 2, 3],
[ 4, 5, 6, 7],
[ 8, 9, 10, 11]])
In [21]: arr2
Out[21]:
array([[12, 13, 14, 15],
[16, 17, 18, 19],
[20, 21, 22, 23]])
In [22]: arr1 + arr2 #对应每个元素相加
Out[22]:
array([[12, 14, 16, 18],
[20, 22, 24, 26],
[28, 30, 32, 34]])
In [24]: arr1 * arr2#对应每个元素相乘
Out[24]:
array([[ 0, 13, 28, 45],
[ 64, 85, 108, 133],
[160, 189, 220, 253]])
In [25]: arr1 + 2 #每个元素都加2
Out[25]:
array([[ 2, 3, 4, 5],
[ 6, 7, 8, 9],
[10, 11, 12, 13]])
In [26]: arr1 / 2 #每个元素都除2
Out[26]:
array([[0. , 0.5, 1. , 1.5],
[2. , 2.5, 3. , 3.5],
[4. , 4.5, 5. , 5.5]])若想要实现矩阵的乘法需要借助于numpy的dot()方法或者使用@作为中间操作符（两个矩阵要满足乘法要求）
In [28]: arr1 = np.arange(6).reshape(2,3)
In [29]: arr2 = np.arange(6).reshape(3,2)
In [30]: np.dot(arr1,arr2)
Out[30]:
array([[10, 13],
[28, 40]])numpy支持同尺寸数组之间的比较，会返回一个布尔值数组
In [38]: arr1 = np.random.randint(0,6,(2,3))
In [39]: arr2 = np.random.randint(0,6,(2,3))
In [40]: arr1
Out[40]:
array([[2, 4, 0],
[4, 3, 2]])
In [41]: arr2
Out[41]:
array([[5, 2, 1],
[0, 0, 2]])
In [42]: arr1 > arr2
Out[42]:
array([[False, True, False],
[ True, True, False]])
In [43]: arr1 > 5#标量比较会将每个元素要进行比较
Out[43]:
array([[False, False, False],
[False, False, False]])对于numpy数组的切片方法与Python中列表切片方法基本一致，但是值得注意的是numpy的数组切片是原数组的视图，换句话说，修改切片会修改原数组
In [51]: arr = np.arange(10)
In [52]: arr
Out[52]: array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
In [53]: ls = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
In [54]: arr_s = arr[:5]
In [55]: arr_s
Out[55]: array([0, 1, 2, 3, 4])
In [56]: arr_s[0] = 10 #改变数组切片会改变原数组
In [57]: arr
Out[57]: array([10, 1, 2, 3, 4, 5, 6, 7, 8, 9])
In [58]: ls_s = ls[:5]
In [59]: ls_s
Out[59]: [0, 1, 2, 3, 4]
In [60]: ls_s[0] = 10#改变列表切片不会改变原数组
In [61]: ls
Out[61]: [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]同理在对切片赋值时，原数组的整个切片区域都将被赋值
In [84]: arr = np.arange(12).reshape(2,2,3)
In [85]: arr
Out[85]:
array([[[ 0, 1, 2],
[ 3, 4, 5]],
[[ 6, 7, 8],
[ 9, 10, 11]]])
In [86]: arr[0,:,1:]#对于多维数组的切片更灵活，但是思想与列表一致
Out[86]:
array([[1, 2],
[4, 5]])
In [87]: arr[0,:,1:] = 250#整个区域都将被赋值
In [88]: arr
Out[88]:
array([[[ 0, 250, 250],
[ 3, 250, 250]],
[[ 6, 7, 8],
[ 9, 10, 11]]])所以在数组中切片要想得到的是数组的拷贝需要用道copy（）方法
In [68]: arr = np.arange(10)
In [69]: arr
Out[69]: array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
In [70]: arr_copy = arr[:].copy()#使用copy方法返回的是原数组的备份，两者是独立存在的
In [71]: arr_copy
Out[71]: array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
In [72]: arr_copy[0] = 10
In [73]: arr
Out[73]: array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
In [74]: arr_copy
Out[74]: array([10, 1, 2, 3, 4, 5, 6, 7, 8, 9])在数组中一个比较方便的操作是可以直接进行布尔索引（生成的是数组拷贝，原数组并没有任何变化）
array([[[ 0, 250, 250],
[ 3, 250, 250]],
[[ 6, 7, 8],
[ 9, 10, 11]]])
In [89]: arr = np.random.randn(4,5)
In [90]: nums = np.arange(4)
In [91]: arr
Out[91]:
array([[ 0.96815772, 1.09588192, -0.75565206, -0.33541785, -1.55058099],
[ 0.79433089, 1.87963211, 0.28907563, 1.41863177, -1.20095493],
[ 2.21006144, -0.80173754, 1.30744596, 0.06783745, 1.01946617],
[ 0.61298953, -0.94047895, -0.63993907, -0.63451676, 1.63448396]])
In [94]: arr[nums%2==0]#0,2
Out[94]:
array([[ 0.96815772, 1.09588192, -0.75565206, -0.33541785, -1.55058099],
[ 2.21006144, -0.80173754, 1.30744596, 0.06783745, 1.01946617]])利用布尔索引操作数组相当灵活
同时与别与列表的常规索引，numpy数组可以实现神奇索引，简单来说可以通过传递一个包含指明所需顺序的列表或数组来完成，如下。
In [2]: arr = np.arange(24).reshape(6,4)
In [3]: arr
Out[3]:
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 [4]: arr[[0,2,4]]#传递一个列表获取数据
Out[4]:
array([[ 0, 1, 2, 3],
[ 8, 9, 10, 11],
[16, 17, 18, 19]])
In [6]: arr[[0,2,4],[1,2,3]]#传递一个列表获取数据
Out[6]: array([ 1, 10, 19])