Sorry I should have proofread a bit better. Or what about using this as a third term, like y[i] = w1[i]*m1 + w2[i]*m2 + > w12[i]+m12 + b[i]? >
I meant w12[i]*m12 for the third term. One way to add patterns to the classifier might be to have input vectors for > 3x3 patterns. Instead of a 1 at the location of all the stones in the 3x3 > pattern you could have some small value, and zero elsewhere > This is a mistake. It obvious doesn't matter whether the pattern entries are all 1's and zeros or all 0.1's and zeros, since the weight vector will take care of their contribution. The only time we care is if we are weighting some entries more than others, like when I suggested we have an input term with 1 at the last move and 0.x < 1 at the second-to-last move. > And this is where offline learning from game records could come in handy, > for initializing the p[i]'s, etc. > I should have said we can learn the initial weights for the pattern inputs from game records.
_______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
