[Python] Is this a bug in NumPy histogram function?

This Python 3.12.7 script with numpy 2.2.4:

import numpy as np

a = np.random.randint(0, 256, (500, 500)).astype(np.uint8)
counts, bins = np.histogram(a, range(0, 255, 25))
print(np.column_stack((counts, bins[:-1], bins[1:])))
counts, bins = np.histogram(a, range(0, 257, 16))
print(np.column_stack((counts, bins[:-1], bins[1:])))


produces this kind of output:

[[24721 0 25]
[24287 25 50]
[24413 50 75]
[24441 75 100]
[24664 100 125]
[24390 125 150]
[24488 150 175]
[24355 175 200]
[24167 200 225]
[25282 225 250]]
[[15800 0 16]
[15691 16 32]
[15640 32 48]
[15514 48 64]
[15732 64 80]
[15506 80 96]
[15823 96 112]
[15724 112 128]
[15629 128 144]
[15681 144 160]
[15661 160 176]
[15558 176 192]
[15526 192 208]
[15469 208 224]
[15772 224 240]
[15274 240 256]]


where the first histogram always has the highest count in bin [225, 250). The second histogram indicates a uniform distribution, as expected. I tried a dozen of times and the anomaly was always there. Can someone explain this behavior?

Continue reading...