Off Topic > Off Topic
Making a new evolution simulator.
DrenDran:
--- Quote from: Nonnel on February 13, 2017, 02:47:41 AM ---Very neat. Looks like editing the creatures' actual code (genome?) is pretty advanced.
--- End quote ---
Thanks, yeah, it takes some getting used to. I have a google doc that contains all the details needed to know how to reprogram cells.
--- Quote from: Nonnel on February 13, 2017, 02:47:41 AM ---How big have you seen the organisms get?
--- End quote ---
They can get reasonably large, I haven't added protein food yet so there's no real downside to having a larger surface area. I'd say maybe 10-15 parts is common.
--- Quote from: Nonnel on February 13, 2017, 02:47:41 AM ---Is it usually the case that one organism is totally dominant and the others die out/don't expand or is there usually a good competition?
--- End quote ---
If you leave it long running long enough with no barriers eventually you do get a dominant species. That said, if you add barriers you can get completely different sets of species. Removing the barriers afterwards produces an environment where multiple species can compete for food for sometimes hours before one finally emerges dominant.
--- Quote from: Mega-Bear on February 13, 2017, 12:31:56 PM ---The diversity of cells in this video is very interesting. I think you may have created real life here.
--- End quote ---
Thanks!
--- Quote from: Meldaril on February 13, 2017, 05:38:37 PM ---If I'm understanding correctly, a global minima in this program would be a solution that is the optimal "eater" and "reproducer". How is a canididate solution represented using chromosomes/genetic sequences? This isn't a clear-cut optimisation problem and I'm struggling to understand the method you are going to use for selection. I imagine you'll tweak the mutation rate to combat local minima.
Convergence is impossible without crossover so it's going to be very interesting to see your program when you have it implemented.
Edit:
I read your earlier post, and you mentioned that there is no fitness function which is why I'm curious about this. A fitness function is a part the selection process needed for crossover.
--- End quote ---
I've seen and examined plenty of programs that intend to use genetic algorithms (neural networks seem to be pretty hip these days as well) to accomplish a specific task like moving as fast a possible, or forming a picture picture, or playing video games or whatever. This isn't that. This is meant to exhibit evolution as it is in real life, the result of completely random things happening that just happen to establish order and patterns. There is no 'goal' but producing as many eggs as possible and all other functions the creature can do (seeing, mating, moving around, fighting) are optional.
DrenDran:
The whole thing with having body parts able to be damaged and removed has caused so many bugs.
Trying to get it to work I basically forgeted up the whole program. It sucks because I haven't made progress this week because nothing I do works.
This weekend I'll have to rewrite a decent amount of the code to try and get it working again. I really should make backups of the code.
DrenDran:
Haven't posted here in a week.
Will probably make new video soon and release an early version of the simulator for other's to try.