Original Post
NN's
http://verveproject.blogspot.com/200...-humanoid.html
Watch and behold
Neural Networks are really not that advanced, and training it can easily be done with genetic algorithms, a "the fittest survive" kind of deal.
I think it would be awesome if some people got together to try and put together an AI with this principle. I'm not gonna try my hands on this until LUA has become a more established part of Toribash (I hope Hampa is working on this =)), but I would like to know if there's guys here interested in this stuff.
If you're interested in this, you really should read
this: http://www.ai-junkie.com/ga/intro/gat1.html
and this: http://www.ai-junkie.com/ann/evolved/nnt1.html
So, is there interest?
Watch and behold
Neural Networks are really not that advanced, and training it can easily be done with genetic algorithms, a "the fittest survive" kind of deal.
I think it would be awesome if some people got together to try and put together an AI with this principle. I'm not gonna try my hands on this until LUA has become a more established part of Toribash (I hope Hampa is working on this =)), but I would like to know if there's guys here interested in this stuff.
If you're interested in this, you really should read
this: http://www.ai-junkie.com/ga/intro/gat1.html
and this: http://www.ai-junkie.com/ann/evolved/nnt1.html
So, is there interest?
Erm, read the Uke-bot threads. It's already done. (Albeit a bit simpler due to not having proper inputs on Uke's part)
Problem is, to use Neural Nets you need I/O. Once you get I/O, it'll be pretty easy.
Also, look for NERO the game. It's a game based around Neural Nets and training them. Just playing it and understanding the mechanics made NN's simple for me. : )
Problem is, to use Neural Nets you need I/O. Once you get I/O, it'll be pretty easy.
Also, look for NERO the game. It's a game based around Neural Nets and training them. Just playing it and understanding the mechanics made NN's simple for me. : )
Originally Posted by SSJokker
Erm, read the Uke-bot threads. It's already done. (Albeit a bit simpler due to not having proper inputs on Uke's part)
Problem is, to use Neural Nets you need I/O. Once you get I/O, it'll be pretty easy.
Also, look for NERO the game. It's a game based around Neural Nets and training them. Just playing it and understanding the mechanics made NN's simple for me. : )
I've seen your bot
I just meant taking it further, but that's not possible yet, as you have too little information (as you said). I'm a NN novice, but they're damn simple (atleast on the surface, I'm sure there's lots of cool algorithms to use/develop)
Oh, and I'll check out that game, it sounds cool
.
I'm taking a course on NN's. Not sure if you guys are doing the math, but I know I am.
Anywho, what are the limitations on commands on multiplayer?
I already know I can't
And another question, are we trying to find? A "pattern" of winning move sets? Secondly, freeze1 and freeze2 are dependant events. How should we treat that?
I started my own AI, without realizing I/O problem. I had a different idea than a real NN.
Yes, I know this is an old post, but AI's will always get some ones attention.
Anywho, what are the limitations on commands on multiplayer?
I already know I can't
for i = 0,19 do
for x,y in pairs(get_joint_info(enemy,i)) doAnd another question, are we trying to find? A "pattern" of winning move sets? Secondly, freeze1 and freeze2 are dependant events. How should we treat that?
I started my own AI, without realizing I/O problem. I had a different idea than a real NN.
Yes, I know this is an old post, but AI's will always get some ones attention.
Last edited by FNugget; Jan 13, 2008 at 05:56 AM.
Ok well I have a little experience with NN but not so much with Tori and the lua system so maybe people can answer some of my questions and we can get something done.
The SDK provides the ability to get information about the world state (such as joint position for either player, damage), so I dont know what SSJokker is talking about when he says there is no I/O: use LUASOCKET!!!!!!!
Luasocket is a lua library which will enable you to send UDP packets back and forth to JNEAT (a very nice and modular java neat implementation) which is what does the heavy lifting for you.
Heres the bottlenecks to getting this thing working as I see it:
NEAT is very hungry for fitness tests. To get a reasonable solution it needs many generations of fitness tests; each generation will have a population ranging from 100s to tens of thousands.
That means with an initial pop. of 1000 controllers, 20 generations in we require over 20000 matches of Toribash to take place.
Getting the information from Toribash to JNEAT isn't the problem here, its getting a reasonable amount of tests done in a very limited timespan.
Another thing occurs to me: NN fail when they are applied to finding a complex solution right off the bat.
To make this work well, the best bet is first teaching the NN something relatively simple, like not falling over, then moving up to walking, then moving up to hurting the enemy. That means we need our fitness tests to become progressively harder once one of the neural nets stumbles upon an optimal solution.
The SDK provides the ability to get information about the world state (such as joint position for either player, damage), so I dont know what SSJokker is talking about when he says there is no I/O: use LUASOCKET!!!!!!!
Luasocket is a lua library which will enable you to send UDP packets back and forth to JNEAT (a very nice and modular java neat implementation) which is what does the heavy lifting for you.
Heres the bottlenecks to getting this thing working as I see it:
NEAT is very hungry for fitness tests. To get a reasonable solution it needs many generations of fitness tests; each generation will have a population ranging from 100s to tens of thousands.
That means with an initial pop. of 1000 controllers, 20 generations in we require over 20000 matches of Toribash to take place.
Getting the information from Toribash to JNEAT isn't the problem here, its getting a reasonable amount of tests done in a very limited timespan.
Another thing occurs to me: NN fail when they are applied to finding a complex solution right off the bat.
To make this work well, the best bet is first teaching the NN something relatively simple, like not falling over, then moving up to walking, then moving up to hurting the enemy. That means we need our fitness tests to become progressively harder once one of the neural nets stumbles upon an optimal solution.
Last edited by tripwire; Feb 27, 2008 at 03:44 AM.
