Hi. I worked on this a long time ago. It's a script that uses a basic Genetic Algorithm configuration (10 population size, and roulette wheel fitness competition for parent selection) to evolve a sequence of move that will get Tori closer to Uke. The Script is a mess, but if you run it and press space to start the GA, you should be able to see incremental improvement.

Thanks!

I've seen it before, I think. Good video. It's amazing stuff. My script isn't even in the same concept of league of what MarI/O. Still fun tho.

Thought you were this guy for a second.
http://forum.toribash.com/showthread.php?t=170100

Aww man, not only is my [CONCEPT] script unoriginal, but also a piece of crap cromparred to the first from half a decade ago... Oh well, it was nice feeling like I contributed, while it lasted.


Thanks for the info!

Hey no worries man - yours looks great. I actually stumbled across the first one because I got interested in making my own genetic algorithm implementation in Toribash (and now I can see it's been done twice!).

I'm still very interested in using Toribash in some kind of AI capacity though - I was thinking machine learning could be fun. But with the amount of game/replay data that you'd need to pull that off there's a whole lot of other awesome projects and analytics that could be done too.

Edit: Something that would be cool to do that, as far as I know, has never been done, would be to try to come up with some fitness functions and try to see if you can use a genetic algorithm to get close to some records.

Daaamn the mario video posted on the first thread was impressive ! It'd be amazing to actually be able to implement AIs in tori !

Those two scripts are equally as impressive ! And I am definitely looking forward to see what you can do in the future, Daanando, if you continue to work and improve your work (And I really hope you will)