On 4/18/13 1:05 PM, glen wrote:
Marcus G. Daniels wrote at 04/18/2013 11:39 AM:
Well, I've done this before on a real problem using a monadic interface
of Bullet physics to Haskell.
Nice.  Is it open?  Or lost in some well of secrecy somewhere?
Um, err, it was a SC11 demo, so it's been shown, but I'd need to fill out paperwork to distribute it.
Meh.

It's a 3D model of enzymatic degradation of cellulose. A couple kinds of enzyme agents, one that breaks hydrogen bonds thereby exposing glycosidic bonds for breaking by another enzyme type. Idea is to find the right mix of enzymes to optimize the rate at which agricultural waste wood can be broken down into gluclose, and then fuel. Spatial considerations like are important in containing the reactions. Going further and doing a full blown molecular dynamics simulation would take a solar system worth of computers. Thus the hybrid approach.

Obviously, the decoupling of agents and environments is not a new idea. Karl Sims '94, etc.
Because the types are an assumed ontology, inferences made by the simulation can be classified into "yeah, we programmed it that way" vs. "hey, that's interesting". The extent to which strong or weak typing matters should depend on the requirements, the use cases for the simulation. I'd like to find a way to choose strong or weak typing depending on what I'm trying to do.
Ok, assertions can be added to check for expected constraints on mixed sets or types can be grown to tolerate some or total mixing. Sort of the same idea. But I now find I prefer the latter because it requires explicit consent at compile time. Tracking down when a errant member in a mixed set appeared in a weakly typed collection requires putting assertions all over the place or intercepting all set insertions and winding the stack back to see who did it. Makes me sad.

Marcus

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
to unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com

Reply via email to