? ????????? ?? 10 ????? 2011 15:13:32 ????? dlakelan ???????:
> On 03/10/2011 02:46 PM, Valery Pipin wrote:
> > Why not to make differentiation in spectral space?
> > I don't see a problem in if maxima will differentiate directrly as well.
> > You will need go to numbers in any case either following to Galerkin
> > procedure, or estimating polynomials in collocation points
> > (pseudospectral)
>
> Yes, I could go to an entirely numerical procedure. In maxima, doing
> collocation points would probably not be too difficult.
>
> My problem is simple enough that if I am willing to wait a long time the
> Galerkin projections can be done exactly via maxima's integrate. I
> thought perhaps if I can keep it from expanding things into basic
> polynomials then I can use the orthogonal properties and much simplify
> the calculation of the Galerkin projection.
>
> I will do numerical procedures next if I can not get a symbolic form for
> the Galerkin projections.
Ok, I have no time to wait. I do Galerkin via Gauss integration(just
summation).
Calculation of matrices for the eigen-value mean-field solar dynamo model
takes about 1day(24H) on 4core 2.53Gz linux mashine. The matrix rank is
500x500.