Re: Delayed Differential Equations [NDelayDSolve]
- To: mathgroup at smc.vnet.net
- Subject: [mg65210] Re: Delayed Differential Equations [NDelayDSolve]
- From: "ben" <benjamin.friedrich at gmail.com>
- Date: Sat, 18 Mar 2006 06:40:32 -0500 (EST)
- References: <dvdibe$a1c$1@smc.vnet.net><dve4p6$ffn$1@smc.vnet.net>
- Sender: owner-wri-mathgroup at wolfram.com
Thanks for your reply, Jens.
I still have questions about your answer:
1) can this still be down in mathematica, or do i have to write/use
some c-program?
2) are there easy ways to solve delayed differential equations for lazy
mathematica
users like me which maybe take a bit longer than a specialized
continuous output initial solver
but still complete their task in reasonable time and don't produce
problems with
precission?
I would define anyting as reasonable fast, if it takes 1-5 times as
long as NDSolve
for the non-delay DE.
Also you said, that NDelayDSolve is slow,
because so is NDSolve.
But integrating my equation without the delay using NDSolve is
reasonable fast.
Actually, i think the problem is that
Timing[NDSolve[...,{t,0,tint}]]
does not scale linearly with tint, but saturates for low values.
Since NDelayDSolve splits the domain of integration up
into tiny pieces of length equal the delay \tau,
the time for NDelayDSolve will diverge as \tau^(-1).
Bye
Ben