Python numpy array math showing weird results

Question

The last cell of the bellow array shows a wrong result:

Code:

import numpy
b=numpy.array ([[1,2,3,4],[5,6,8,16]])
b=b**b/b+1
b

result:

array([[2.000000e+00, 3.000000e+00, 1.000000e+01, 6.500000e+01],
 [6.260000e+02, 7.777000e+03, 2.097153e+06, 1.000000e+00]])

All numbers are correct except for the last one.

Type is numpy.float64. The correct answer is 1.152921504606847e+18.

Any ideas?

Thanks

Answer 1

When you construct a numpy array with whole numbers, its type will be int32 or int64 depending on your system, instead of the float64 you expected. So when you do 16**16, the result overflowed and you end up with a 0.

To fix this, specify the float64 type for your numpy array, like this:

b=numpy.array ([[1,2,3,4],[5,6,8,16]], dtype='float64')

Or if you want, add a .0 to any of the numbers in your array to specify that you want to use float not int.

• Ehsan: To add to it, I think the default is int64 and if you want to check, overflow happens since np.iinfo(np.int64).max/(16**16) is smaller than one.
• juanpa.arrivillaga: Actually, the dtype will be some integer dtype, but the exact size is platform-dependant, so whether it is a 64 bit vs a 32 bit architecture and the OS as well. Basically, it is whatever a C long is. Assuming a 64bit operating system, that is a 64bit integer on Linux, but a 32 bit integer on windows.
• juanpa.arrivillaga: Also, you should just do numpy.array([[1,2,3,4],[5,6,8,16]], dtype=numpy.float64)