[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

computer-go: Hybrid/learning



I was just looking over the code for gnugo, I don't know if its
representative; but there are tons of parameters (values of given moves
primarily, but also various fiddly things buried elsewhere) which seem
arbitrary.

Why not apply a learning approach (eg self-play to find the exact best
values of these?

Josh