> -----Original Message----- > From: Jason House <[EMAIL PROTECTED]> > To: computer-go <computer-go@computer-go.org> > Sent: Thu, 13 Dec 2007 3:20 pm > Subject: Re: [computer-go] MC-UCT and tactical information
> On Dec 13, 2007 2:17 PM, <[EMAIL PROTECTED]> wrote: I'd like to start a more specific discussion about ways to combine tactical information with MC-UCT. Here's the scenario. It's the bot's turn and, prior to starting any playouts, it runs a tactical analyzer (for want of a better name) that labels each string as unconditionally alive, unconditionally dead, conditionally alive, uncertain, or too strong to bother testing. For each conditionally alive string, it finds the move(s) to kill/rescue it. Each possible move, has a critical score indicating the net number of stones whose life/death status would be changed for the better if?the bot?moved there.?Moves with a non-zero critical score are "critical moves." > As worded, I don't know if I agree with that formalization of what > information will be useful.? > It's in the right ballpark for the discussion, but will likely shift a bit > with any final design.? ???? Right. It's not meant to be comprehensive. By... um?coincidence, that's the information my bot has access to. I'm soliciting ideas on what?other information?should be included as well as how to use it. It's also a hint at the minimum level of detail I'm hoping for. ?... > What's an n-gram? ???? The?idea would be to use n-grams in much the same way as they are used in machine language translation.?Wikipedia has good descriptions for both. To give the flavor of it, an n-gram is a sequence of length n. In machine translation, they are usually sequences of words. In this case, it would be a sequence of?moves. I have a body of examples, the local searches,?and I calculate a score for each short sequence of moves that came up. Now, in my global playouts, when I encounter one of these sequences again, I have a measure of goodness for the next move in that sequence. This would be a way of trying to generalize between different lines of play. ??? A move could be just the coordinates of the stone, or it could include pattern information. or whatever. What I *don't* want to do is wind up storing?every huge hairy hash table that came up in the local searches. - Dave Hillis ? ________________________________________________________________________ More new features than ever. Check out the new AIM(R) Mail ! - http://webmail.aim.com
_______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
