OdesolveMinerr Mathcad 11

[jeanGiraud]

This collection will be augmented soon

Original work sheet updated 20100827

jmG

_______________________________

Substantially augmented [jmG 2010830]

[ViktorKorobov]

Thanks Jean! Very interesting work sheet. It works very quickly and much more precisely.

Mathcad 11

Obsolete way

calc_time=52.9 s, ERR^2=1.97e-06 (odesolve/fixed, 5000 steps)

calc_time=118.1 s, ERR^2=1.97e-06 (odesolve/adaptive, 5000 steps)

calc_time=0.2 s, ERR^2=1.64e-06 (odesolve/stiff, 5000 steps)

Your way

calc_time=0.9 s, ERR^2=7.35e-07 (odesolve/fixed, 10000 steps)

calc_time=2.1 s, ERR^2=7.35e-07 (odesolve/adaptive, 10000 steps)

calc_time=0.1 s, ERR^2=1.20e-06 (odesolve/stiff, 10000 steps)

It is a pity that this file does not work in 14 versions.

Viktor

[AlanStevens]

Nice neat method jean!

It is trivially modified to work in M14/15 - see attached. Still about twice as fast in M11 though!

Alan

[ViktorKorobov]

We rejoiced prematurely. Jean's and Alan's worksheets do not work. Try to change initial guesses and look result!

Viktor

[AlanStevens]

Oops! Just shows how important it is to do some basic checks and not just rely on the fact that it doesn't produce a run-time error.

Back to your method I think Viktor!

Alan

[jeanGiraud]

Viktor Korobov wrote:

We rejoiced prematurely. Jean's and Alan's worksheets do not work. Try to change initial guesses and look result!

Viktor

You are correct Viktor, the constrains in the Minerr block must be inclusive .

100 points is enough [default Odesove], rkfixed, Pentium IV XP in < 1 second.

Viktor, you have no choice about the initial conditions, they are dictated by the data set.

Thanks Alan it you can make it work for versions 14, 15