Custom tolerance tests

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Custom tolerance tests

Joza
Normally, when I write a zero finding algorithm, at stage k, my convergence test tests whether or not the obtained value is less than machine epsilon times the the previous value.

IF   x_k < mach_eps x_(k-1)   then STOP

But if you use say fzero or fsolve, the built-in functions, oen can only set an absolute tolerance. That is, one can only set a value such that the algorithm will stop when it obtains a value less than the specified tolerance. You cannot use the kind of test above, where you test consecutive values.

Is there a way around this that I am not aware of? What would be the best way to specify the tolerance for say, fzero, in line with the tolerance test above?
Reply | Threaded
Open this post in threaded view
|

Re: Custom tolerance tests

eem2314


On Tue, Oct 23, 2012 at 6:06 AM, Joza <[hidden email]> wrote:
Normally, when I write a zero finding algorithm, at stage k, my convergence
test tests whether or not the obtained value is less than machine epsilon
times the the previous value.

IF   x_k < mach_eps x_(k-1)   then STOP

But if you use say fzero or fsolve, the built-in functions, oen can only set
an absolute tolerance. That is, one can only set a value such that the
algorithm will stop when it obtains a value less than the specified
tolerance. You cannot use the kind of test above, where you test consecutive
values.

Is there a way around this that I am not aware of? What would be the best
way to specify the tolerance for say, fzero, in line with the tolerance test
above?



--
View this message in context: http://octave.1599824.n4.nabble.com/Custom-tolerance-tests-tp4645655.html
Sent from the Octave - General mailing list archive at Nabble.com.
_______________________________________________
Help-octave mailing list
[hidden email]
https://mailman.cae.wisc.edu/listinfo/help-octave

I believe you want to use the TolX option (see fsolve documentation)

--
Ed Meyer


_______________________________________________
Help-octave mailing list
[hidden email]
https://mailman.cae.wisc.edu/listinfo/help-octave