Suppose I have 1x1x1x1x… array and wish to convert it to scalar?

How do I do it?

squeeze does not help.

import numpy as np

matrix = np.array([[1]])
s = np.squeeze(matrix)
print type(s)
print s

matrix = [[1]]
print type(s)
print s

s = 1
print type(s)
print s

You can use the item() function:

import numpy as np

matrix = np.array([[[[7]]]])
print(matrix.item())

Output

7

Numpy has a function explicitly for this purpose: asscalar

>>> np.asscalar(np.array([24]))
24

This uses item() in the implementation.

I guess asscalar was added to more explicit about what’s going on.

You can index with the empty tuple after squeezing:

x = np.array([[[1]]])
s = np.squeeze(x)  # or s = x.reshape(())
val = s[()]
print val, type(val)

You can use np.take

np.take(matrix,0)

Sample run –

In [15]: matrix = np.array([[67]])

In [16]: np.take(matrix,0)
Out[16]: 67

In [17]: type(np.take(matrix,0))
Out[17]: numpy.int64

>>> float(np.array([[[[1]]]]))
1.

You can use the numpy.flatten() function to flatten the array to a one dimensional array.
For instance:

s = [[1]]
print(s[0][0])
>>>> 1

or

a=[[1],[2]]
print(a[0][0])
>>>> 1
print(a[1][0])
>>>> 2