swarm-modeling
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Swarm-Modelling] Re: Non-integer timesteps


From: dgokaydin
Subject: Re: [Swarm-Modelling] Re: Non-integer timesteps
Date: Tue, 26 Apr 2005 21:23:42 +0100 (WEST)
User-agent: SquirrelMail/1.4.4

Hello and thank you for your replies,

First of all, please note that you are dealing with someone little or no
experience in computation, let alone mathematics. This said, let me try to
explain to you my view of things.

My only interest in modelling non-integer timesteps is avoiding some kind
of bias or computational aberration resulting from event superimposition.
If two events are supposed to have an order in my model I want them to
happen in that order, not just be scheduled to the next timestep and then
randomly, or in a fixed sequence, happen.

Avoiding event superimposition wouldn't necessarily require floating point
numbers, or non-integers, as I stated in my first post. This is because,
conceptually, floating point numbers truncated to have a fixed number of
decimals, as all in computation must be, are as discrete as integers. Just
multiply them by the necessary power of 10 to make them integers, then
send them to the schedule. That was my first thought, but then I decided
to ask this distinct pannel of experts because I imagined that playing
with schedules would do the trick, saving me a lot of time wasted with a
not very elegant solution to the problem. I still haven't seen the
examples that you so kindly provided, due to absolute shortage of time.
But I most certainly will.

>From my unenlightened perspective, I suppose that you are absolutely right
about being impossible to (at least fully) implement continuous time
models in computation. But hey, we do it all the time! Every number is
stored in a computer with a limited number of bytes right? Everytime I
solve an equation in my computer those are discrete timesteps.

In practice, and just so that you understand my problem more objectively,
I want to model a Poisson process with constant rate. The time to my next
event is thus drawn from an exponential distribution with mean equal to
the rate with which my event happens. You can understand I won't get any
integers doing this. :-)

Dinis Gökaydin



reply via email to

[Prev in Thread] Current Thread [Next in Thread]