# Script adapted into a funtion Function not returning equal values

3 views (last 30 days)
C Meek on 22 Jul 2012
Hey everyone,
I'm having a problem where two identical pieces of code are returning different values when run as a script then a function. Trying to keep this as simple as possible, I have two vectors Z and Zt, which represent curves with a number of prominent peaks. I find the peak locations of each curve, then compare their separation distance using an adapted Gaussian function:
1 - exp( (-mu0*Phi(ii)) / (2 * vp^2));
This is an extract from a loop, where Phi contains the difference between each peak of Z and Zt. mu0 and vp won't particularly matter here.
In particular, I have Z and Zt set so that their peak locations are in exactly the same place - and as a result the expression above will be
1 - exp(0) = 0;
for all entries in Phi.
This is exaclty what I want, but when I run this with a function call, it is returning very small values of order 10^-5, instead of zero.
Anyone have any ideas of why this may be?

Geoff on 22 Jul 2012
Floating point precision errors. If you need to test for zero, then test for something like:
if abs(value) < 1e-5
10^-5 does sound rather large for a zero value though.. Are you using single-precision floats?
C Meek on 23 Jul 2012
I'm using double precision. Turns out that the unexpected small numbers were coming from another part of my code which I hadn't taken into account - so all is well, when I adapt it according zero is returned ok!
Thanks