Abstract
This paper deals with a learning mechanism which uses the statistical decision method. It discusses in detail the minimax decision function which is usually effective when the control experience is limited. It deduces the necessary condition for the minimax decision function in the light of the theory of games. It also shows that the optimal minimax decision function is obtainable by the linear programming method.
The paper goes on touch on the learning ability of the optimal minimax decision function. It suggests the relaxed minimax decision function obtained by relaxing the condition for the minimax decision function. The function thus obtained is suggested for use where the optimal minimax decision function lacks some learning ability.
The relaxed minimax decision function, the paper tells us, does the same work that the minimax decision function does where the control experience is limited. As the control experience increases, the function can do the same work that the empirical Bayes decision function does.
