numpy subtract two arrays: output

Question

So I want to subtract two numpy arrays a and b:

a=[[ 1. 0.85979163 0. 0.11766047 0.19353699]
[ 0.8589698 1. 0.24111901 0. 0. ]
[ 0. 0.24554123 1. 0.09234979 0.07125199]
[ 0.31269982 0.22558714 0.29298401 1. 0.475543 ]
[ 0.18880995 0. 0.06580817 0.32276821 1. ]]


b=[[1 1 1 1 1]
[1 1 1 1 1]
[1 1 1 1 1]
[1 1 1 1 1]
[1 1 1 1 1]]

When I use the following command:

y=numpy.subtract(b,a)

I get an output array:

y= [[ -2.22044605e-16 1.40208370e-01 1.00000000e+00 8.82339528e-01
8.06463005e-01]
[ 1.41030195e-01 0.00000000e+00 7.58880995e-01 1.00000000e+00
1.00000000e+00]
[ 1.00000000e+00 7.54458767e-01 -2.22044605e-16 9.07650211e-01
9.28748013e-01]
[ 6.87300178e-01 7.74412855e-01 7.07015986e-01 0.00000000e+00
5.24457002e-01]
[ 8.11190053e-01 1.00000000e+00 9.34191829e-01 6.77231787e-01
0.00000000e+00]]

And it's really confusing me as to how it's outputting these values.

When I tried to troubleshoot and do:

y=b[0,0]-a[0,0]

I got y=0.. which makes sense because i would be subtracting 1-1. But in the output array I'm getting a value of -2.22044605e-16 instead.

Do you have any idea as to why this might be happening and what I can do to fix it?

I cannot reproduce your error with the data that you provided (I get all 0s on the main diagonal). I hypothesize that `a` is not exactly as you describe it. Perhaps it is a result of another computation, and `1.0` is not exactly `1.0`. — DYZ, Sep 24 '18 at 01:04
There is no "fixing it" it is the nature of representing numbers in a computer. See https://en.wikipedia.org/wiki/Floating-point_arithmetic — Matt, Sep 24 '18 at 01:12
Possible duplicate of [how set numpy floating point accuracy?](https://stackoverflow.com/questions/25181642/how-set-numpy-floating-point-accuracy) — Matt, Sep 24 '18 at 01:13

Julien · Answer 1 · 2018-09-24T01:06:23.853

e-16 is essentially 0. You typically can't avoid numerical precision errors when dealing with floats. But you can always chop the tiny values to 0 exactly afterwards if that's important to you. Note that you probably see 1. when printing a only because numpy doesn't show the full precision by default when printing an array. Try print(a[0,0]) and you'll probably see something like 1.00000000005. For example try printing np.array([1.000000001]) and np.array([1.000000001])-1, and np.array([1.000000001])[0]...

numpy subtract two arrays: output

1 Answers1