| 摘要 |
PURPOSE:To automatically execute an optimal control at the time of copy in accordance with the respective diverse states of various documents by using a neural network. CONSTITUTION:A neural network is constituted of an input layer LI, a middle layer LM and an output layer LO, and each unit is coupled with each unit of other layer than the layer to which it belongs through a coupling coefficient. To the unit of the input layer LI, the minimum value, an average value and a selection mode value which are read by an automatic exposure sensor are inputted successively at the time of pre-scan, and in accordance with synapse weight of the network, a calculation is executed in order of the input layer the middle layer the output layer by prescribed learning and a control value such as an exposure value, etc., is outputted. Data of a coupling coefficient obtained by comparing a result of calculation with a teacher signal and correcting an error so as to become minimum, and the part for executing an output calculation are integrated into the device body, and with respect to plural input conditions, a calculation is executed at the time of necessity and an output condition (control condition) is outputted. In such a way, the memory capacity can be curtailed, and also, copying can be executed with an optimal picture quality. |
