PURPOSE: To provide an image forming device which can execute the time division sharing of one network by a large amount of control, and realize careful control in a comparatively small-sized network, when an optimum input/ output relation can be constructed and simultaneously, executed by the learning of some patterns that is, in a neural network algorithm actually mounted on a product.

CONSTITUTION: Detecting means 11-13 with respect to many environmental parameters are provided, they are inputted into an input layer, and a signal for optimizing a control quantity to an image quality adjusting means is turned into an output signal from an output layer. During this time, a nerve cell model on which the intermediate layer of a proper stage is set, is used, and a weight constant is determined, in a system in which all output units to determine the control quantity having optimum balance and size, with respect to all adjusting means are included in the output layer. On the other hand, at the time of executing, the neural network with respect to a signal adjusting means is shared by the time division by each adjusting means.