You may want to use the ndarray.item method, as in a.item(). This is also equivalent to (the now deprecated) np.asscalar(a). This has the benefit of working in situations with views and superfluous axes, while the above solutions will currently break. For example,
>>> a = np.asarray(1).view()
>>> a.item() # correct
1
>>> a[0] # breaks
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
IndexError: too many indices for array
>>> a = np.asarray([[2]])
>>> a.item() # correct
2
>>> a[0] # bad result
array([2])
This also has the benefit of throwing an exception if the array is not a singleton, while the a[0] approach will silently proceed (which may lead to bugs sneaking through undetected).
>>> a = np.asarray([1, 2])
>>> a[0] # silently proceeds
1
>>> a.item() # detects incorrect size
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
ValueError: can only convert an array of size 1 to a Python scalar