[SciPy-Dev] Adding startiter keyword to scipy.integrate.quadrature

Andrew Nelson andyfaff@gmail....
Sat Apr 27 12:15:42 CDT 2013


Dear dev list,
please forgive any glaring faux pas I might make.  This is my first
time contributing to scipy development.

I am proposing to add the 'startiter' keyword to scipy.integrate.quadrature.

At the moment the adaptive quadrature starts from a Gaussian
quadrature order of 1 and finishes when maxiter iterations are reached
(unless the termination criteria are met).

For the problem I deal with I know that I have to start iteration from
approx. order 15 onwards. Which means that orders 1 through 14 are
wasted. This would correspond to 105 calculations of the function to
be integrated (1+2+3+4...+14).

Adding the startiter keyword enables one to start at an arbitrary
order, saving a reasonable amount of calculation time.

Would such this addition be suitable for inclusion into scipy?  If so,
I can prepare a pull request.

regards,
Andrew.

On 24 April 2013 13:13, Andrew Nelson <andyfaff@gmail.com> wrote:
> Dear dev list,
> please forgive any glaring faux pas I might make.  This is my first
> time contributing to scipy development.
>
> I am proposing to add the 'startiter' keyword to scipy.integrate.quadrature.
>
> At the moment the adaptive quadrature starts from a Gaussian
> quadrature order of 1 and finishes when maxiter iterations are reached
> (unless the termination criteria are met).
>
> For the problem I deal with I know that I have to start iteration from
> approx. order 15 onwards. Which means that orders 1 through 14 are
> wasted. This would correspond to 105 calculations of the function to
> be integrated (1+2+3+4...+14).
>
> Adding the startiter keyword enables one to start at an arbitrary
> order, saving a reasonable amount of calculation time.
>
> Would such this addition be suitable for inclusion into scipy?  If so,
> I can prepare a pull request.
>
> regards,
> Andrew.
>
>
>
> --
> _____________________________________
> Dr. Andrew Nelson
>
>
> _____________________________________



-- 
_____________________________________
Dr. Andrew Nelson


_____________________________________


More information about the SciPy-Dev mailing list