It is difficult for von Neumann-type computer to read letters and understand conversation. The neural network is researched to make information processing system which is difficult to achieve with von Neumann-type computer.
In this study, we evaluate the board in Othello game by using perceptron which is a kind of neural network. Board evaluation is to express the advantageous of board by the numerical value.
I tried to supervised learning to the network by using 500,000 game records which was existing Othello program fought mutually.
Two methods were used to input board status into the network. The one is to divide the stones on the board into the black stones and the white stones and to input them to the network. The other is to pick up several characteristics from board and to make index patterns by using it.
As a result of the former's method, I found that the error was decreased by learning. But it wasn't enough to use in practical. In comparison with the method of the former, the latter has shown better results.
How to decrease the error in the last stage is future assignment.