如何创建一个仅包含具有特定形状的 inf 的数组？

Question

这是一个 Matlab 向量：a = [inf(m,1);ones(m,1)]

我尝试用 python 方式创建一个类似的对象。

我试过这个：

import numpy as np

a = np.stack((np.tile(np.array([np.inf]), (m, 0))),np.ones((m,0)))

例如，通过在控制台中将 m 更改为 5 并尝试使用 PyCharm 查看它，我收到了以下消息：

Traceback (most recent call last):
  File "C:\Program Files\JetBrains\PyCharm 2019.1.3\helpers\pydev\_pydev_comm\server.py", line 34, in handle
    self.processor.process(iprot, oprot)
  File "C:\Program Files\JetBrains\PyCharm 2019.1.3\helpers\third_party\thriftpy\_shaded_thriftpy\thrift.py", line 266, in process
    self.handle_exception(e, result)
  File "C:\Program Files\JetBrains\PyCharm 2019.1.3\helpers\third_party\thriftpy\_shaded_thriftpy\thrift.py", line 254, in handle_exception
    raise e
  File "C:\Program Files\JetBrains\PyCharm 2019.1.3\helpers\third_party\thriftpy\_shaded_thriftpy\thrift.py", line 263, in process
    result.success = call()
  File "C:\Program Files\JetBrains\PyCharm 2019.1.3\helpers\third_party\thriftpy\_shaded_thriftpy\thrift.py", line 228, in call
    return f(*(args.__dict__[k] for k in api_args))
  File "C:\Program Files\JetBrains\PyCharm 2019.1.3\helpers\pydev\_pydev_bundle\pydev_console_utils.py", line 359, in getArray
    return pydevd_thrift.table_like_struct_to_thrift_struct(array, name, roffset, coffset, rows, cols, format)
  File "C:\Program Files\JetBrains\PyCharm 2019.1.3\helpers\pydev\_pydevd_bundle\pydevd_thrift.py", line 602, in table_like_struct_to_thrift_struct
    return TYPE_TO_THRIFT_STRUCT_CONVERTERS[type_name](array, name, roffset, coffset, rows, cols, format)
  File "C:\Program Files\JetBrains\PyCharm 2019.1.3\helpers\pydev\_pydevd_bundle\pydevd_thrift.py", line 377, in array_to_thrift_struct
    array, array_chunk, r, c, f = array_to_meta_thrift_struct(array, name, format)
  File "C:\Program Files\JetBrains\PyCharm 2019.1.3\helpers\pydev\_pydevd_bundle\pydevd_thrift.py", line 476, in array_to_meta_thrift_struct
    bounds = (array.min(), array.max())
  File "C:\Users\Azerty\PycharmProjects\OptionsHedgeFund\venv37\lib\site-packages\numpy\core\_methods.py", line 32, in _amin
    return umr_minimum(a, axis, None, out, keepdims, initial)
ValueError: zero-size array to reduction operation minimum which has no identity

这有什么问题？

Answer 1

Numpy 有专门用于填充数组的函数。请参阅“完整（）”，“填充（）”。

a = np.full((2, 2), np.inf)
array([[inf, inf],
       [inf, inf]])

Answer 2

在 Octave 会话中：

>> m = 5;
>> a = [inf(m,1); ones(m,1)];
>> size(a)
ans =

   10    1

>> a
a =

   Inf
   Inf
   Inf
   Inf
   Inf
     1
     1
     1
     1
     1

在ipython 会话中：

In [21]: m=5                                                                                                 
In [22]: np.vstack([np.ones((m,1))*np.inf, np.ones((m,1))])                                                  
Out[22]: 
array([[inf],
       [inf],
       [inf],
       [inf],
       [inf],
       [ 1.],
       [ 1.],
       [ 1.],
       [ 1.],
       [ 1.]])
In [23]: _.shape                                                                                             
Out[23]: (10, 1)

做同样事情的变体：`

np.concatenate([np.full((m,1), np.inf), np.ones((m,1))], axis=0)

===

对于两行，从 (1,m) 形状开始：

>> a = [inf(1,m); ones(1,m)];
>> size(a)
ans =

   2   5

>> a
a =

   Inf   Inf   Inf   Inf   Inf
     1     1     1     1     1

In [26]: np.concatenate([np.full((1,m), np.inf), np.ones((1,m))], axis=0)                                    
Out[26]: 
array([[inf, inf, inf, inf, inf],
       [ 1.,  1.,  1.,  1.,  1.]])

===

至于您的错误，np.stack 在连接两个 (5,0) 形状的阵列时遇到问题（我不完全确定原因）。

In [27]: a = np.stack((np.tile(np.array([np.inf]), (m, 0))),np.ones((m,0)))                                  
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-27-c6f8ec8462c9> in <module>
----> 1 a = np.stack((np.tile(np.array([np.inf]), (m, 0))),np.ones((m,0)))

/usr/local/lib/python3.6/dist-packages/numpy/core/shape_base.py in stack(arrays, axis, out)
    417 
    418     result_ndim = arrays[0].ndim + 1
--> 419     axis = normalize_axis_index(axis, result_ndim)
    420 
    421     sl = (slice(None),) * axis + (_nx.newaxis,)

TypeError: only size-1 arrays can be converted to Python scalars

你的错误是不同的；看起来它是由pydev 生成的，而不是 Python 本身。

但是检查一下：

In [28]: np.tile(np.array([np.inf]), (m, 0))                                                                 
Out[28]: array([], shape=(5, 0), dtype=float64)
In [29]: np.ones((m,0))                                                                                      
Out[29]: array([], shape=(5, 0), dtype=float64)

In [31]: a = np.vstack([(np.tile(np.array([np.inf]), (m, 0))),np.ones((m,0))])                               
In [32]: a                                                                                                   
Out[32]: array([], shape=(10, 0), dtype=float64)

将 (m,0) 替换为 (m,1)，我们得到所需的 (10,1) 数组：

In [33]: a = np.vstack([(np.tile(np.array([np.inf]), (m, 1))),np.ones((m,1))]) 

stack 添加一个维度 - 这不是您想要的：

In [35]: a = np.stack([(np.tile(np.array([np.inf]), (m, 1))),np.ones((m,1))])                                
In [36]: a.shape                                                                                             
Out[36]: (2, 5, 1)

虽然有一个简单的起始形状，但我们得到 2 行：

In [37]: a = np.stack([(np.tile(np.array([np.inf]), (m,))),np.ones((m))])                                    
In [38]: a.shape                                                                                             
Out[38]: (2, 5)

实际上，您的stack 的问题是错误的)。这是正确的：

In [50]: np.stack((np.tile(np.array([np.inf]), (m, 0)),np.ones((m,0))))                                      
Out[50]: array([], shape=(2, 5, 0), dtype=float64)

您的结束 ) 在 tile 之后，因此 np.ones() 表达式位于 axis 参数位置。

Answer 3

因此，如果我正确理解了您的预期结果，您可能会这样做来生成矩阵。

m = 5 #or whatever

a = np.ones((2, m)) #create a matrix 2 x m of ones
a[0] = np.inf #replace first row with infinites

a 是：

array([[inf, inf, inf, inf, inf],
       [ 1.,  1.,  1.,  1.,  1.]])