computer-go: Neural networks

["Julian Churchill" <jjc97c@xxxxxxxxxxx>]

Hi,

 I have just started a project to incorporate neural network techniques into
a standard alpha beta search algorithm. I am wondering if anyone else has
tried to apply neural networks to the domain of Go? I read about a program
called NeuroGo a while back, it appears to be in version 2 now, but no new
info.

The actual idea is to use a neural network trained on some professional
games, as a sort of ultra fast pattern matcher, but it would also pick up
trends and general techniques to apply to playing Go as it was trained,
that's one of the features of neural networks. I am thinking of using this
to suggest plausible moves given the current board state and then these
could be fed into a minimax search, this could be taken further where the
network suggests moves for each level in the minimax tree, keeping it
focused. If the network was good enough at suggesting appropriate moves then
only a few at each level need be considered, say a branching factor of 6. I
hope this will be a useful improvement to the minimax search, especially
combined with pruning techniques, allowing a much deeper search.

Any thoughts about this would be greatly appreciated.

 Cheers,
 Jules

 Julian Churchill