PURPOSE: To provide a method and device for controlling the feedback of a process by using a neural network.

CONSTITUTION: A desired process output value is outputted. A practical process output value corresponding to the required process output value is measured. Then both the output values are mutually compared. A practically measured process output value expressing maximum variation from the corresponding desired process output value is determined based upon respective compared results. The practical process output value expressing the maximum deviation from the corresponding desired process output value is controlled to generate a predictive process output value. The controlled process output value is inputted to the neural network trained so as to provide the reverse model of the process. The neural network generates a signal for driving the process in response to the predictive process value and the process generates at least one output value the most approximate to the corresponding desired process output value.