Author Topic: field-population size question  (Read 2908 times)

Offline viplex

  • Bot Neophyte
  • *
  • Posts: 45
    • View Profile
field-population size question
« on: January 31, 2006, 02:48:43 PM »
Hi

My computer is a little slow. I would like to see something interesting evolving from a short bot. All my efforts have been a failure, for the DNA I get is always a heap of (seemingly) useless trash, or maybe I just cannot w8 for a good result.
My question is: should I set field and population size to small, or set it to large? Which one is better for a quicker evolution? I have been thinking on this a lot.   :tantrum:
« Last Edit: January 31, 2006, 02:52:44 PM by viplex »

Offline Numsgil

  • Administrator
  • Bot God
  • *****
  • Posts: 7742
    • View Profile
field-population size question
« Reply #1 on: January 31, 2006, 04:58:24 PM »
Well it depends on exactly what you want to see happen.

Bots aren't really good at evolving new behaviors.  Shot bots rarely learn to tie, etc.  Instead, bots tend to get better at the whatever it is they already do.  Shot bots learn to always shoot.  Bots in high friction sims learn to compensate, things like that.

A population size of about 100 bots should be enough to geta  good sim going.  As to field size, at present it shouldn't make a huge difference, but a size of about 3 is best I think.

Half the trick is patience and the other half is being able to figure out success when you see it.  Evolution has a really chaotic effect on DNA, so looking at DNA itself don't be expecting too much.

Offline viplex

  • Bot Neophyte
  • *
  • Posts: 45
    • View Profile
field-population size question
« Reply #2 on: February 01, 2006, 03:29:32 PM »
Alas, yes, I tend to realize that. No new tricky behaviour. I have thought of something that I havent tried yet though, filling DNA with LOADS of useless stuff myself, using many commands and variable names without use, so that evolution may put them in use some time later. This may result in higher variance & quiker evolveing for my Celeron 733 am I right.
 Viplex

Offline shvarz

  • Bot God
  • *****
  • Posts: 1341
    • View Profile
field-population size question
« Reply #3 on: February 01, 2006, 05:02:39 PM »
You are right, the more stuff you have, the better are the chances of something new develop.

It is also important to create conditions that force the bot to evolve.  If it's doing quite good already, then the selective pressure is low and evolution proceeds slowly.  You have to play with environment to create a situation where you have a stable population of bots that are barely surviving.  It's tricky, but it can be done.
"Never underestimate the power of stupid things in big numbers" - Serious Sam

Offline Endy

  • Bot Overlord
  • ****
  • Posts: 852
    • View Profile
field-population size question
« Reply #4 on: February 01, 2006, 10:35:51 PM »
I play with the physics settings myself. Make tricky worlds for the bots to evolve in, encouraging dna to find ways to take advantage of the situation. High Gravity or brownian have been the most interesting so far.

I've started playing with encoding currently useless stuff into the bots like:

50 .sharenrg *.multi mult store

Which gives non-sharefeeding bots, the ability to begin share feeding as soon as they start mutating  their birth-tie deletion tie.

I'm planning on adding a fairly comprehensive wikki on all the different behaviors I've seen develope so far(a pretty large list actually B) ). Should be a good resource for everyone.
« Last Edit: February 01, 2006, 10:36:47 PM by Endy »