To make the observables of a process participate in common to each other in decision of the output value by performing the optimum correlating of an execution function via a function approximating unit based on the probability distribution of all value to which the probability distribution of the output value is supplied.

An input vector consists of xt, xt-1, xt-2, yt, yt-1 and yt-2. The vectors xt and yt show the present value, and the vectors xt-1, xt-2, yt-1 and yt-2 show the past value. Then the process observables are simulated by a function approximating unit and only the past value of all time series become available via the function approximating unit. Thus every value of the time series is generated in a proper probability distribution, and the probability process is carried out and grasped. Then the value generated by the function approximating unit is added to every present value of the time series of the output value. At the same time, the function approximating unit generates a function as an execution function to guarantee that the optimum correlating is secured in the probability distribution of all value to which the probability distribution of the output value is supplied.