通用函数（ufunc），是一种在 ndarray 数据中进行逐元素操作的函数。

一元通用函数

In [1]: import numpy as np

In [2]: arr = np.arange(10)

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

In [4]: np.sqrt(arr)
Out[4]:
array([0.        , 1.        , 1.41421356, 1.73205081, 2.        ,
       2.23606798, 2.44948974, 2.64575131, 2.82842712, 3.        ])

In [5]: np.exp(arr)
Out[5]:
array([1.00000000e+00, 2.71828183e+00, 7.38905610e+00, 2.00855369e+01,
       5.45981500e+01, 1.48413159e+02, 4.03428793e+02, 1.09663316e+03,
       2.98095799e+03, 8.10308393e+03])

In [6]:

二元通用函数

In [6]: x = np.random.randn(8)

In [7]: y = np.randoom.randn(8)
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-7-2458dd362115> in <module>
----> 1 y = np.randoom.randn(8)

AttributeError: module 'numpy' has no attribute 'randoom'

In [8]: y = np.random.randn(8)

In [9]: x
Out[9]:
array([-2.18256212, -0.08144067, -0.08418644, -0.01270041, -0.78696876,
       -0.46798034, -0.22818082, -0.69275325])

In [10]: y
Out[10]:
array([ 0.69679623,  1.73440592,  0.96581309, -0.12895321, -1.30429304,
        1.23067988,  0.45521522,  0.60378768])

In [11]: np.maximum(x, y)
Out[11]:
array([ 0.69679623,  1.73440592,  0.96581309, -0.01270041, -0.78696876,
        1.23067988,  0.45521522,  0.60378768])

numpy.maximum逐个元素地将 x
和 y 中元素的最大值计算出来。

一些 ufunc 返回多个数组

In [12]: arr = np.random.randn(7) * 5

In [13]: arr
Out[13]:
array([ 4.06149543, -0.66663476, -0.95013597, -2.28401568, -2.44863962,
        3.16878977,  3.11914078])

In [14]: remainder, whole_part = np.modf(arr)

In [15]: remainder
Out[15]:
array([ 0.06149543, -0.66663476, -0.95013597, -0.28401568, -0.44863962,
        0.16878977,  0.11914078])

In [16]: whole_part
Out[16]: array([ 4., -0., -0., -2., -2.,  3.,  3.])

modf 返回一个浮点值数组的小数部分和整数部分。

通用函数接受一个可选参数 out 。

In [35]: arr
Out[35]:
array([2.01531522,        nan,        nan,        nan,        nan,
       1.78010948, 1.76610894])

In [36]: arr2 = np.zeros(7)

In [37]: np.sqrt(arr, arr2)
Out[37]:
array([1.41961798,        nan,        nan,        nan,        nan,
       1.33420744, 1.32895031])

In [38]: arr2
Out[38]:
array([1.41961798,        nan,        nan,        nan,        nan,
       1.33420744, 1.32895031])