On Nov 13, 2007 11:31 AM, Ian Osgood <[EMAIL PROTECTED]> wrote: > > On Nov 13, 2007, at 7:46 AM, Jason House wrote: > > > > > > > On Nov 13, 2007 10:36 AM, Ian Osgood <[EMAIL PROTECTED]> wrote: > > I like Forth. I got excited about UCT around the time of the Computer > > Olympiad and wrote a bitmap-based 9x9 program. What is the general > > impression on bitmap vs. mailbox board representations for Monte > > Carlo readouts? > > > > > > I never went down the road of bitmap based go because I had not > > clever way to efficiently track captures. How did you get around > > this hurdle? > > > > I never claimed efficiency. > > What do you mean by "track captures"? You mean detect when a string > loses all its liberties? That is simply when the intersection of the > set of empty points with the string expanded once is the empty set. > This can only happen to strings bordering a move, so I do the check > there:
That shifts the problem to detecting the strings that border a move. It seems to imply needed a non-bit-board with string data to work (ie. an indication of what string is at each board location). Without tracking something like that, string detection becomes the looping operation.
_______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
