Different result for the same mathematical operation

Question

I'm using Python3 and numpy and I'm doing the following:

diff1 = np.abs(self.trainX - sample_repeated)
diff2 = np.abs(sample_repeated - self.trainX)
print("diff1 equals diff2?")
if(np.array_equal(diff1,diff2)): 
    print("Yes")
else:
    print("No")

I expected diff1 to be equal to diff2. Any idea why the code always prints "No"? All the arrays have the same same (self.trainX, sample_repeated, diff1 and diff2).

NB: numpy docs report that array_equal is "True if two arrays have the same shape and elements, False otherwise."

Solved: you arrays was uint8 and that leads to overflow. You must cast them to int before

Possible duplicate of [Is floating point math broken?](https://stackoverflow.com/questions/588004/is-floating-point-math-broken) — user2390182, Sep 29 '17 at 22:16
Floating points always have inaccuracies, so it is never advisable to compare with == or similar. — eyllanesc, Sep 29 '17 at 22:18
I have checked and all the arrays written in the code are uint8. What if it tries to do the sum 255 + 255? Overflow? — user1315621, Sep 29 '17 at 22:28
Everybody take notice: this question is about **uint8**, not float! — Mark Ransom, Sep 29 '17 at 22:32

score 0 · Accepted Answer · answered Sep 29 '17 at 22:31

0

Solved: you arrays was uint8 and that leads to overflow. You must cast them to int before

answered Sep 29 '17 at 22:31

user1315621

3,044
9
42
86

Different result for the same mathematical operation

1 Answers1