摘要 |
PROBLEM TO BE SOLVED: To easily perform learning of relation between input data and output data as time series data. SOLUTION: Each input network net<SB>in1</SB>, net<SB>in2</SB>in an input/output relation model where nodes of the input networks net<SB>in1</SB>, net<SB>in2</SB>and output networks net<SB>out1</SB>, net<SB>out2</SB>constituted by the plurality of nodes having a time series pattern model expressing a time series pattern are coupled to each other, is updated in a self-organizing manner on the basis of input data. Each output network net<SB>out1</SB>and net<SB>out2</SB>is updated in a self-organizing manner on the basis of the output data. The coupling relation between a winner node of the nodes of the input networks net<SB>in1</SB>, net<SB>in2</SB>and each node of the output network net<SB>out1</SB>, net<SB>out2</SB>is updated, and the coupling relation between the winner node of the nodes of the output network net<SB>out1</SB>, net<SB>out2</SB>and each node of the input network net<SB>in1</SB>, net<SB>in2</SB>is updated. This device, method and program can be applied to, for example, a robot. COPYRIGHT: (C)2007,JPO&INPIT
|