MATLAB cannot round non-zero entries to a specific decimal place

Asked Mar 02 '18 at 03:46

Active Mar 02 '18 at 07:22

Viewed 32 times

What does the following statement mean?

"MATLAB cannot round non-zero entries to specific decimal place, because of that there will always be leftovers around 1e-16 for doubles (double precision variables) and 1e-12 for singles."

EXAMPLE:

time(3001)=0 and find(time==0)=3001
but time(3000)=-1e4 and find(time==-1e4) = []
because time(3000) + 1e-4 = -2.11e-16 and single(time(3000)) + 1e-4 = 2.53e-12.

edited Mar 02 '18 at 07:22

Sardar Usama

19,536
9
36
58

asked Mar 02 '18 at 03:46

Mussie Tesfay

4

I think the note message is pretty clear and they even provide a example! – scriptboy Mar 02 '18 at 03:49
What does it mean by leftovers? – Mussie Tesfay Mar 02 '18 at 03:52
Read this to know how computer store float: https://en.wikipedia.org/wiki/Single-precision_floating-point_format . – scriptboy Mar 02 '18 at 03:58
2

For example, compute `0.0000001 + 0.0000000001` you will get `1.0009999999999999e-07` not ``0.0000001001` – scriptboy Mar 02 '18 at 04:02
So this does not only apply to Matlab? – Mussie Tesfay Mar 03 '18 at 13:03
@MussleTesfay every single programming language. It all deals with how the CPU deals with floating point numbers as they're not infinite precision. I also recommend reading the perils of floating point for a good picture too: http://www.lahey.com/float.htm – rayryeng Mar 03 '18 at 14:35
Very helpful indeed. Thank you so much! – Mussie Tesfay Mar 03 '18 at 16:21

MATLAB cannot round non-zero entries to a specific decimal place

0 Answers0