| |
| |
| |
|
Page: 1 2
Comments:
<0> <JohnnyL@ef> anyone here use opencyc?
<1> i couldn't g et it to work on 64bit though
<0> <JohnnyL@ef> cwnner, same problem as me.
<0> <JohnnyL@ef> it ****s.
<2> pardon me?
<0> <JohnnyL@ef> waspdigger: what?
<2> oh nothing
<0> <JohnnyL@ef> cwenner, have you posted on opencyc the problem that you have?
<3> hi
<0> <some_pers@ef> Wow! this is very active!
<0> <some_pers@ef> so it looks
<0> <some_pers@ef> any bots here?
<0> <some_pers@ef> did someone do something to boot me?
<0> <some_pers@ef> jsut curious
<0> <some_pers@ef> what;s that?
<0> <some_pers@ef> *what's
<0> <some_pers@ef> is somethign supposed to happen?
<2> what?
<0> <some_pers@ef> anyway's what's mode/#ai [+1 34] by CHAThost?
<0> <some_pers@ef> *anways,
<0> <some_pers@ef> anyways, what's a good chatterbot?
<2> search the net
<0> <some_pers@ef> no recommendations?
<3> try the ones that have AIML.
<0> <some_pers@ef> AIML?
<0> <some_pers@ef> *uses wikipedia* I see
<0> <some_pers@ef> anyways, who has heard of "elbot"?
<4> anyone specialize in python in here?
<4> hello
<5> moin
<6> hi
<7> hallo
<8> hey guys anyone understands about minimax?
<9> Yes, I do.
<10> i know a thing or two
<10> hi Ceran
<9> hey chessguy
<8> ok thanks. I am having a problem ussually function of minimax returns a number
<10> generally, yes
<8> thats is how i see from the implementations and from the psedo-code
<9> right
<8> i need it to return a board
<8> like an object
<9> you don't, really
<8> well i cannot do nothing with a number
<9> sure you can
<8> what can i do ?
<9> You calculate the minimax value for each action, then you pick the action with the highest if your maximizing, (or lowest if you're minimizing.
<8> so lets say that i generate all sons of a board
<8> then i calculate minimax for each one of them
<9> remember, minimax just "looks ahead" to determine how the actions in the current board state play out. As long as you know what actions are currently available, all you need from minimax is a number.
<8> so i need to know which actions are available and for each i am calculating the minimax right?
<9> correct
<8> Ceran: i see
<8> Ceran: thanks
<8> Ceran: actually i am implementing it with alfa beta cut seems more efficient
<9> Yes, AB is the useful form of minimax.
<11> as long as you have an accurate, efficient heuristic
<8> so you can say that its something like this -> http://pastebin.ca/333350
<9> I've unsuccessfully tried implementing MTDF and Negascout.
<8> rahul: it has a little ramdom bias to garate that never generates two equal evaluations
<11> you don't want random bias
<9> iturk: btw, it's Alpha, not Alfa
<11> you want to actually find the best move.
<8> Ceran: well its portuguese style sorry
<12> can someone please explain how kernel tricks work to me?
<8> rahul: its really little
<11> iturk: even then, you don't want to make a wrong move.
<8> rahul: its ok i am using a genetic algorigthm
<11> heh
<11> like that'll help if you only allow it to generate incorrect algos
<8> rahul: it will help to make some mutations
<9> iturk: equal evaluations are fine
<11> iturk: no. mutations help to make mutations.
<9> You don't want "mutations". AB is deterministic.
<9> iturk: What domain are you using this for?
<8> Ceren that part its not inside the AB algorithm
<8> Ceran: AI
<8> Ceran: its used to optimize the evaluation function return
<9> iturk: I meant, tic-tac-toe, chess, shoots-and-latters? what are you using it for?
<8> chess
<9> ah
<8> actually its a new aproach no one tried before
<8> its called PSO
<8> i mean in chess games
<9> Yeah, Negascout is the current standard for most chess programs, which is a variation of alphabeta.
<8> particle swarm optimization
<8> Ceran: i have heard about it
<8> Ceran: deepblue was using it right ?
<9> iturk: I have a working AlphaBeta implementation with transposition table.
<9> I'm not sure what DeepBlue used.
<11> it was alpha-beta
<9> chessguy can probably help you there ;)
<8> Ceran: the computer that plays chess
<10> chessguy hasn't been paying attention
<9> I know what DeepBlue is :P
<8> Ceran: ;P
<8> Ceran: actually i have never heard about transposition tables
<10> they're also frequently called hash tables in chess engines
<9> iturk: A transposition table is just a cache that AB uses to store repeated calculations.
<13> it is used to prune loops right?
<10> ydo, it's used to prevent re-searching positions which have already been searched
<9> Yes, most programming languages refer to them as hash tables as well.
<10> iturk, i've done some research on applying genetic programming to chess.
<8> Ceran: hash tables its more familiar to me
<8> chessguy: this is more or less other area its called co-evolutionary particle swarm optimization
<9> I just called them transposition tables because that's usually what they're referred to in literature about the alphabeta algorithm.
<10> oh, a swarm algorithm
<8> chessguy: i am having problems because i dont have cluster with me
<8> chessguy: so my computer its really getting hot
<10> Ceran, in the chess domain the term 'transposition' actually makes sense, because even chess players who don't program refer to transpositions in variations
<10> iturk, is there a more detailed description of your approach somewhere?
<8> swarm intelligence its the key
<9> but swarms are stupid :)
<8> Ceran: yes they are
<10> i've done some reading about swarm theory, and a lot about chess AI, but i've seen little on a sensible approach to tying the two together
<8> Ceran: but they can behave like intelligent guys if they are many
<10> iturk, more info please :)
<8> chessguy: i have a bible here wait
<8> PSO-based coevolutionary game learning by cornelis J. Franken
<10> cheers
<8> he has applied swarm with success to checkers
<10> how successful have you been with chess so far?
<9> Well, here's a working ABwMemory, written in Python, in case you'd like a reference: http://pastebin.ca/333384
<8> chessguy: i have search a lot i havent find anyone trying to apply this to chess. if you find tell me
<8> chessguy: till now the evaluation its very poor
<8> chessguy: only considering material
<10> it took me about 8 or 10 months before i finally found a single paper on applying GP to chess
<8> chessguy: i hope you will find my after some months if you try to find by PSO :P
<10> iturk, have you written a typical alpha-beta searcher for chess?
<8> chessguy: http://pastebin.ca/333350
<10> in java? no wonder your machine is melting :)
<8> chessguy: is it depth 10 enough or i should go further
<8> Ceran: python its really a eficient language.
<10> that depends on a lot of things
<8> chessguy: i am using it static
<10> what kind of board representation are you using?
<10> (by the way)
<8> chessguy: element[][]
<10> ewww
<8> char element[][]
<8> chessguy: its really melting but unless its portable
<10> you've got some reading to do :)
<10> http://www.cis.uab.edu/hyatt/pubs.html is a good place to start
<8> chessguy: i see
<8> chessguy: well till now memory its not a problem
<8> chessguy: and most of the time its lost in the PSO iteractions
<10> it's not a matter of memory
<8> chessguy: you just mean time
<10> other board representations give you much faster ways to determine what pieces are where, and generate moves, etc.
Return to
#ai
or
Go to some related
logs:
bind dig returns nothing
#openzaurus
load deamon xscreensaver
apt-get remove xserver-xorg-core
#bash
ubuntu guarddog shortcut
#linux
script scandir for .exe
mp4creator rpm fc5
wireless-tools 28pre9 ubuntu
|
|