Thanks, interesting book! I downloaded it completely, oh my poor finger.


Smitty:

I see how kd-trees help in finding the nearest neighbor etc., but at the moment I don't see a use for them to solve the problem explained in my first post. I have to align objects to be neighbors, not test if/who they are. Or did you look already further than I?

lol what is this,

I printed those 12 pages of the one pdf,
and seemingly all words which have it lack of the character combination "fi",
ie. "defined" is printed as "de ned", "configuration" as "con guration"


is my printer/driver dumb, or is this a kind of nasty "copy protection", so that I can't just print-out the book instead of buying it ?
It's good that I didn't print more chapters . . . argh.

In the Numerical Recipes book, it says the strategy also involves to occasionally acceppt "uphill steps", ie. steps that make the current solution worse instead of better.
I see the need for that, but what I'm not sure about:

What are good strategies to decide when to accept an "uphill" step ?

My implementation is working so far, and until now pretty good, but I'm afraid after adding a couple more restrictions(penalties) it won't do that good without having an uphill step here and there, which isn't there at all currently.

[Edited by - UnshavenBastard on April 6, 2006 2:50:19 AM]