[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: archives and general heatbug question
From: |
Nelson Minar |
Subject: |
Re: archives and general heatbug question |
Date: |
Wed, 27 Mar 96 13:56:38 MST |
>First trying to access the latest archives on web page
>(list-archive.9602 list-archive.9603) generates a "Forbidden You don't
>have permission to access..." message but the others are fine.
yeah, that's a bug in our SFI installation. You can get the files via
majordomo, or I go in occasionally and fix it (like now).
>How about converting the archives with a hypermail program (something like
>mhonarc) that will generate a index of all of the messages for each archive?
Great idea, would take some time to set up.
>Second. Has anyone played around with different values in heatbugs?
Yes, heatbugs is great fun. It's a nice demo for a variety of reasons.
It lets us tell the "local behaviour, global emergent phenomena" story
in a nice way. That, and 64 reds look impressive on a projector :-)
>For example try just changing the diffusion constant to above 1... say 1.5.
>A checkerboard like pattern appears in our version and the "bugs"
>always move in a linear fashion from the right to the left often
>forming streams of bugs along light grey strings that form a
>boundary/diffuse barrier.
That's degenerate behaviour of the diffusion equation. Looks cool, but
not physical. The source code for Diffuse2d includes a comment on how
diffusion is approximated by the CA. Here it is:
// basic CA diffusion.
// newHeat = evap (self + k * (nbdavg - self))
// where self is our own heat
// nbdavg is the appropriately weighted average of our neighbour's heat
// k is the diffusion constant
// evap is the evaporation rate.
Values of k above 1.0 are some sort of odd positive feedback loop.
There might be overflow happening, too, I didn't test this case very much.
>Try changing the evaporation rate to above 0.99... say 1.0. In this case
>it appears that the bugs will undergo a "phase change". On our system it
>is around time step 300 to time step 600.
Evaporation rates of 1.0 cause heat to not leave: above 1.0, it
actually adds heat. When the world is saturated with heat, there's not
much in the way of gradient for the bugs.
>Interesting behavior. Is it the demo or the implementation of the demo?
That's the $64 question for simulation, in all cases. With heatbugs I
think I can explain and justify most of its behaviour, so I'm
satisfied. (The big outstanding bug is that heatbugs have a bias in
movement: if two choices are equally acceptable, it goes with the
first one it looks for instead of a random one.)