I have noticed that with an array of completely random numbers, using np.sum vs using a for-loop to iteratively add values seems to produce different values. I have provided the code below.

I want to assume there is some floating-point number related innacuracies going on, but I can’t see why this would be the case/ One thing I have attempted is to truncate/round to a finite number of decimal places, but this also seems to trigger a similar problem.

Python 3.9.12
Numpy 1.21.5
Windows 10 Pro

Thanks in advance.

import numpy as np

np.random.seed(1)
x = np.round( np.random.uniform(0, 1, 100), 3)

y = np.sum(x.copy())
z = 0.0
for xi in x.copy():
    z += xi

print("y", y)       # With seed=1, y=48.585
print("z", z)       # With seed=1, z=48.585000000000015
print(y-z)

I have tried truncating/rounding to a finite number of decimal places.