i'm having a little problem wit a tic tac toe game i wrote using a bac-propagation neural network as the opponent (A.I) the thing is, the network has to train itself with each board state i.e whenever i make a move, it trains itself using the current state of the board as inputs (i used the following rules. 1 means the point is occupied by an O, -1 means its occupied by an X (the human player) while zero stands for an empty cell. My question is, since this training takes time will it be possible for me to just train the network once using all possible board configurations? currently i train once AND still train each time i make a move.
b.t.w my current training set is about 500 lines of code like this :
0, 1, -1
1, 0, 0
1, -1, 0
i'll be really grateful if anyone could help out with this problem.