Neural network device

Abstract

Abstract PURPOSE:To speed up learning speed, and to improve probability to converge to an optimum solution in a wide view by constituting a device so as to minimize locally energy defined by output data and the square of calculated output. CONSTITUTION:The device is constituted so that input data and the output data are set beforehand in the neural element 11a of an input layer and the neural element 11c of an output layer respectively as learning data, and the operating state of the neural element 11a of the input layer is allotted actually equally in the same group, and the neural element 11b of an intermediate layer obtains the calculated output at the neural element 11c of the output layer while it total-calculates a signal through a coupled element 12 coupling with the neural element 11b of a preceding layer respectively, and minimizes locally the energy defined by the output data and the calculated output. Thus, the learning speed can be speeded up, and the probability to converge to the optimum solution in a wide view can be improved.

Inventors

• 👤 Yoshio Izui
• 👤 良夫 泉井

Bibliographic Details