A learning system independent from specific models is valuable in engineering. Such system is expected to acheve versatile motor learning ability which we can see in cerebellum. Recently, a new approach to artificial recurrent neural network design called Reservoir Computing (RC) was proposed. It came to regarded as a paradigm to realize the learning abiliity and became a vivid research field. In our research, we built a cognition system of object motion in two-dimensional motion patterns and time series prediction. We combined ESN with an error feedback system, and successfully enhanced the prediction accuracy. Finally, we confirm the learning ability of the system through computer simulations.
各種検索
サポート
ヘルプ