PURPOSE: To increase the learning speed to raise a probability of convergence to a globally optimum solution by calculating the total sum of feedback signals and independent signals by each neural element in a neural element group, and setting an aggregate to a minimum energy state while using the total sum value to change the output of the neural element with respect to probability.

CONSTITUTION: Each neural element 11 in the neural element group calculates the total sum of feedback signals passing respective coupling element groups and independent signals which are inputted independently of these feedback signal and are practically equal to one another in each group. The calculated total sum value is used to change the output of the neural element 11 with respect to probability, thereby constituting the aggregate in the minimum energy state. Thus, a globally optimum solution is easily found and the convergence speed is increased.