On Dec 5, 2007 4:44 AM, Lars <[EMAIL PROTECTED]> wrote: > 2. I had run the algorithm on 400 games (including handicap-games) from > the same game-records source Remi used (Section 3.2), but i used an > other month. I concidered only 3x3 shape-patterns and simple non-shape > pattern including dist-to-boarder, dist-to-prevMove, atari, capture, > extension.. > After 1 iterations of the algorithm, I got strenght-parameter values > completely different to the results of the paper (Table 1). Most of my > parameters (including all the dist-to-boarder parameters!) have values > less than 1.0. Does anyone have some explanations on this? > After 5 iteration it is even worse. Most of the low values (less than > 1.0) gets even lower, the high values even higher.
Some of Remi's features will score every move on the board. When this occurs, any scalar multiplication of those feature values will work. This seems to apply to both "distance to move" features and the MC Owner feature. When I get around to doing this (hopefully soon), I'll probably leave out something to better pin the values down. Essentially, omissions are like forcing a value of 1.0. As far as what the values should be in absolute terms, I can't yet comment :( > Anyone of you have similar or other experiences with the algorithm? The holidays are slowing me down. I hope to be ready to compare notes in January. I'm really glad to see someone else doing this. Any lessons learned will be incredibly helpful...
_______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
