Re: Better recursive method?
- To: mathgroup at smc.vnet.net
- Subject: [mg72835] Re: Better recursive method?
- From: "Roman" <rschmied at gmail.com>
- Date: Sun, 21 Jan 2007 06:30:16 -0500 (EST)
- References: <eosh0f$9bs$1@smc.vnet.net>
Hi nandan,
I think the problem is that your pi() and tau() functions are evaluated
analytically if you give analytic expressions for theta, and this takes
large amounts of time. If you look at ByteCount[pi[i,PI/4]], this seems
to grow exponentially with i, which is always a bad sign.
Fix this by asking for S1[N[PI/4]] for instance, while n also is
globally assigned a definite value.
Roman.
nandan wrote:
> Hallo:
>
> I have following recursive functions:
> pi(i, theta) := (2n-1)/(n-1) cos(theta) pi(i-1, theta) - n/(n-1)
> pi(i-2,theta)
> tau(i,theta) := n pi(i, theta) - (n+1) pi(i-1,theta)
>
> with intial values:
> pi(0,theta) := 0
> pi(1,theta) := 1
>
> These should calculate the following function:
> S1(theta) := sum(i=1-nstop) (2i+1)/(i(i+1)) (a(i) pi(i,theta) - b(i)
> tau(i,theta)) (where nstop=40)
>
> This should be pretty easy to calculate for Mathematica. I have written
> it the following way:
> pi[0,theta] := 0;
> pi[1,theta] := 1;
> pi[i_, theta_] := pi[i, theta] =(2n-1)/(n-1) Cos[theta] pi[i-1, theta]
> - n/(n-1) pi[i-2,theta]
> tau[i_,theta_] := n pi[i, theta] - (n+1) pi[i-1,theta]
> S1Temp[i_,theta_] := (2i+1)/(i(i+1)) (a[i] pi[i,theta] - b[i]
> tau[i,theta])
> S1(theta) := Sum[ S1Temp[i,theta], {i, 1, nstop}]
>
> Well, it works fine, but it takes enormous time. I have introduced
> recursive function to save time for calculation of Legendre Polynomials
> of which the function 'pi' and 'tau' consist of. But the same function
> if I write with 2 loops each for 'theta' and 'i' in IDL, it calculates
> in fraction of a second. Even the simple calculation of 'pi[40,pi/4]'
> took more than 350 seconds.
>
> What could be the problem in my logic. Is there any other better way to
> write recursive functions in Mathematica?
>
> Any help will be appreciated!
>
> Regards,
> nandan