Title:
INFORMATION PROCESSING DEVICE AND PROGRAM
Document Type and Number:
WIPO Patent Application WO/2023/135984
Kind Code:
A1
Abstract:
The present invention comprises: a machine learning module unit that is provided with an input layer, a reservoir layer, and a readout layer; and a machine learning processing unit that sets the readout layer of the machine learning module unit in a trained state so that the value at time t of at least one physical quantity that evolves over time in a system associated with a plurality of physical phenomena is input and so that the value of the at least one physical quantity at time t+Δt that differs by a time Δt from the time t is output, wherein the machine learning module unit set in the trained state is provided for predetermined information processing related to the system associated with the plurality of physical phenomena.
More Like This:
Inventors:
KOTANI KIYOSHI (JP)
WATANABE MAKOTO (JP)
JIMBO YASUHIKO (JP)
WATANABE MAKOTO (JP)
JIMBO YASUHIKO (JP)
Application Number:
PCT/JP2022/044915
Publication Date:
July 20, 2023
Filing Date:
December 06, 2022
Export Citation:
Assignee:
UNIV TOKYO (JP)
International Classes:
G06N3/044; G02F1/137; G06N3/049; G06N3/08; G06N20/00
Foreign References:
JP2021531529A | 2021-11-18 |
Other References:
TANAKA, GOHEI ET AL.: "DS-2-3 Echo State Networks with Consideration for Heterogeneity of Neuron Attributes", PROCEEDINGS OF THE 2017 IEICE GENERAL CONFERENCE; 2017.03.22-25, 2017, JP, pages S23 - S24, XP009548178
Attorney, Agent or Firm:
TAKEI Nobutoshi (JP)
Download PDF: