Python笔记：详解使用内置函数创建ndarray-白红宇

Python笔记：详解使用内置函数创建ndarray

阅读量：4250 次

发布时间：2019-05-26

本文共 6917 字，大约阅读时间需要 23 分钟。

引入

import numpy as np

在此引入一次，下面直接使用 np

使用 `np.zeros(shape)` 创建

X = np.zeros((3,4))print()print('X = \n', X) print()print('X has dimensions:', X.shape)print('X is an object of type:', type(X))print('The elements in X are of type:', X.dtype)

输出：

X =[[ 0. 0. 0. 0.] [ 0. 0. 0. 0.] [ 0. 0. 0. 0.]]X has dimensions: (3, 4)X is an object of type: class 'numpy.ndarray'The elements in X are of type: float64

np.zeros(shape) 会创建一个全是 0 并且为给定形状的 ndarray

同 np.array() 一样，内置函数也可以创建指定类型的 ndarray, 参考前文

使用 `np.ones(shape)` 创建

X = np.ones((3,2), dtype = np.int64) # 此处指定类型print()print('X = \n', X)print()print('X has dimensions:', X.shape)print('X is an object of type:', type(X))print('The elements in X are of type:', X.dtype)

输出：

X =  [[1 1] [1 1] [1 1]]X has dimensions: (3, 2)X is an object of type: 
   
    The elements in X are of type: int64

此处我们创建了一个具有指定形状的 ndarray，其中的元素全是 1

使用 `np.full(shape, constant value)` 创建

X = np.full((2,3), 5) print()print('X = \n', X)print()print('X has dimensions:', X.shape)print('X is an object of type:', type(X))print('The elements in X are of type:', X.dtype)

输出：

X =[[5 5 5] [5 5 5]]X has dimensions: (2, 3)X is an object of type: class 'numpy.ndarray'The elements in X are of type: int64

第一个参数是你要创建的 ndarray 的形状，第二个参数是你要向数组中填充的常数值。

常数值这里如果你填写5.0，那么类型自动变为 float64 类型

当然，还可再加第三个参数，比如： dtype = np.int64 , 来更改数据类型

使用 `np.eye(N)` 创建 `单位矩阵`

X = np.eye(5)print()print('X = \n', X)print()print('X has dimensions:', X.shape)print('X is an object of type:', type(X))print('The elements in X are of type:', X.dtype)

输出：

X =[[ 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.]]X has dimensions: (5, 5)X is an object of type: class 'numpy.ndarray'The elements in X are of type: float64

单位矩阵是主对角线上全是 1，其他位置全是 0 的方形矩阵。

线性代数中的基本数组是单位矩阵。

使用 `np.diag()` 创建 `对角矩阵`

X = np.diag([10,20,30,50])print()print('X = \n', X)print()

输出为：

X =[[10 0 0 0] [ 0 20 0 0] [ 0 0 30 0] [ 0 0 0 50]]

使用 `np.arange(start,stop,step)` 创建在给定区间内值均匀分布的 ndarray

x = np.arange(1,14,3)print()print('x = ', x)print()print('x has dimensions:', x.shape)print('x is an object of type:', type(x))print('The elements in x are of type:', x.dtype)

输出：

x = [ 1 4 7 10 13]x has dimensions: (5,)x is an object of type: class 'numpy.ndarray'The elements in x are of type: int64

可以传入一个参数、两个参数或三个参数。

如果只传入一个参数，np.arange(N) 将创建一个秩为 1 的 ndarray，其中包含从 0 到 N - 1 的连续整数。

注意，如果我希望数组具有介于 0 到 9 之间的整数，则需要将 N 设为 10，而不是将 N 设为 9。

均匀分布的数字将包括 start 数字，但是不包括 stop 数字。

step 表示两个相邻值之间的差。允许间隔为非整数，例如 0.3，但是由于浮点数精度有限，输出通常不一致。

因此，如果需要非整数间隔，通常建议使用函数 np.linspace(start, stop, N)。

使用 `np.linspace(start,stop,step)` 创建在给定区间内值均匀分布的 ndarray

x = np.linspace(0,25,10)print()print('x = \n', x)print()print('x has dimensions:', x.shape)print('x is an object of type:', type(x))print('The elements in x are of type:', x.dtype)

输出：

x =  [ 0.          2.77777778  5.55555556  8.33333333 11.11111111 13.88888889 16.66666667 19.44444444 22.22222222 25.        ]x has dimensions: (10,)x is an object of type: 
   
    The elements in x are of type: float64

np.linspace(start, stop, N) 函数返回 N 个在闭区间 [start, stop] 内均匀分布的数字, 即 start 和 stop 值都包括在内。

也可以不包含区间的结束点（就像 np.arange() 函数一样），方法是在 np.linspace() 函数中将关键字 endpoint 设为 False 。

x = np.linspace(0,25,10, endpoint = False)

使用 `np.reshape(ndarray, new_shape)` 创建秩为 2 的任何形状 ndarray

x = np.arange(20)print()print('Original x = ', x)print()x = np.reshape(x, (4,5))print()print('Reshaped x = \n', x)print()print('x has dimensions:', x.shape)print('x is an object of type:', type(x))print('The elements in x are of type:', x.dtype)

输出：

Original x = [ 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19]Reshaped x =[[ 0 1 2 3 4] [ 5 6 7 8 9] [10 11 12 13 14] [15 16 17 18 19]]x has dimensions: (4, 5)x is an object of type: class 'numpy.ndarray'The elements in x are of type: int64

我们将 np.arange() 和 np.linspace() 这些函数与 np.reshape() 函数相结合，创建秩为 2 的任何形状 ndarray。

np.reshape(ndarray, new_shape) 函数会将给定 ndarray 转换为指定的 new_shape。注意：new_shape 应该与给定 ndarray 中的元素数量保持一致。

某些函数还可以当做方法使用, 比如我们用一行代码实现上述示例中的相同结果: Y = np.arange(20).reshape(4, 5)

注意，new_shape 应该与 ndarray 中的元素数量保持一致。

同样，我们也可以使用 reshape() 与 np.linspace() 创建秩为 2 的数组。

X = np.linspace(0,50,10, endpoint=False).reshape(5,2)print()print('X = \n', X)print()print('X has dimensions:', X.shape)print('X is an object of type:', type(X))print('The elements in X are of type:', X.dtype)

输出：

X = [[ 0. 5.] [ 10. 15.] [ 20. 25.] [ 30. 35.] [ 40. 45.]]X has dimensions: (5, 2)X is an object of type: class 'numpy.ndarray' The elements in X are of type: float64

创建随机 ndarray

使用 np.random.random(shape) 函数创建具有给定形状的 ndarray，包含位于半开区间 [0.0, 1.0) 内的随机浮点数。

X = np.random.random((3,3))print()print('X = \n', X)print()print('X has dimensions:', X.shape)print('X is an object of type:', type(X))print('The elements in x are of type:', X.dtype)

输出：

X =[[ 0.12379926 0.52943854 0.3443525 ] [ 0.11169547 0.82123909 0.52864397] [ 0.58244133 0.21980803 0.69026858]]X has dimensions: (3, 3)X is an object of type: class 'numpy.ndarray' The elements in X are of type: float64

使用np.random.randint(start, stop, size = shape)创建由特定区间内的随机整数构成的 ndarray, 其中包含在半开区间 [start, stop) 内的随机整数。

X = np.random.randint(4,15,size=(3,2))   print()   print('X = \n', X)   print()   print('X has dimensions:', X.shape)   print('X is an object of type:', type(X))   print('The elements in X are of type:', X.dtype)

输出：

X =   	[[ 7 11]   	 [ 9 11]   	 [ 6 7]]   	   X has dimensions: (3, 2)   X is an object of type: class 'numpy.ndarray' The elements in X are of type: int64

使用 np.random.normal(mean, standard deviation, size=shape) 创建由满足特定统计学特性的随机数字组成的 ndarray

X = np.random.normal(0, 0.1, size=(1000,1000))print()print('X = \n', X)print()print('X has dimensions:', X.shape)print('X is an object of type:', type(X))print('The elements in X are of type:', X.dtype)print('The elements in X have a mean of:', X.mean())print('The maximum value in X is:', X.max())print('The minimum value in X is:', X.min())print('X has', (X < 0).sum(), 'negative numbers')print('X has', (X > 0).sum(), 'positive numbers')

输出：

X =[[ 0.04218614 0.03247225 -0.02936003 ..., 0.01586796 -0.05599115 -0.03630946] [ 0.13879995 -0.01583122 -0.16599967 ..., 0.01859617 -0.08241612 0.09684025] [ 0.14422252 -0.11635985 -0.04550231 ..., -0.09748604 -0.09350044 0.02514799] ..., [-0.10472516 -0.04643974 0.08856722 ..., -0.02096011 -0.02946155 0.12930844] [-0.26596955 0.0829783 0.11032549 ..., -0.14492074 -0.00113646 -0.03566034] [-0.12044482 0.20355356 0.13637195 ..., 0.06047196 -0.04170031 -0.04957684]]X has dimensions: (1000, 1000)X is an object of type: class 'numpy.ndarray' The elements in X are of type: float64The elements in X have a mean of: -0.000121576684405The maximum value in X is: 0.476673923106The minimum value in X is: -0.499114224706 X 具有 500562 个负数 X 具有 499438 个正数

说明：