Configuration and the study manner null of neural network for calling allocation

摘要

PURPOSE: To prepare an assigning neural net efficient and that of good performance by providing an output layer, corresponding to the predicted waiting time of a landing call, as the output layer of a neural net for call assignment, separately from an output layer corresponding to an assigned elevator. ;CONSTITUTION: In the connecting constitution of a neural net, an output layer corresponding to the predicted waiting time of a landing call is provided separately from an output layer corresponding to an assigned elevator (assignment aptitude), that is, only the number of neurons of the output layer is increased in comparison to the conventional constitution of a neural net, and the teaching signals of the assigned elevator and waiting time are simultaneously supplied to the same input pattern for learning. Since two teaching signals of the waiting time and assigned elevator are thus supplied simultaneously for learning. knowledge reflecting both 'answer' and 'process leading to an answer' is automatically padded in each connected weight.;COPYRIGHT: (C)1996,JPO