Title:
SYSTEM
Document Type and Number:
WIPO Patent Application WO/2023/199846
Kind Code:
A1
Abstract:
The present invention provides a system comprising a first-layer agent in a first layer and a plurality of second-layer agents in a second layer of a lower order than the first layer, the agents being disposed in a network. Each of the plurality of second-layer agents executes cooperative learning with the other second-layer agents using collected information, and transmits information generated using a learning result to the first-layer agent. The first-layer agent executes learning using the information received from the plurality of second-layer agents.
Inventors:
SUDA KAZUTO (JP)
ISHIWAKA YUKO (JP)
OWADA MINORU (JP)
YOSHIDA TOMOHIRO (JP)
ISHIWAKA YUKO (JP)
OWADA MINORU (JP)
YOSHIDA TOMOHIRO (JP)
Application Number:
PCT/JP2023/014300
Publication Date:
October 19, 2023
Filing Date:
April 06, 2023
Export Citation:
Assignee:
SOFTBANK CORP (JP)
International Classes:
G06N20/00
Foreign References:
| JP2021140825A | 2021-09-16 |
Other References:
ALEXANDER SASHA VEZHNEVETS; SIMON OSINDERO; TOM SCHAUL; NICOLAS HEESS; MAX JADERBERG; DAVID SILVER; KORAY KAVUKCUOGLU: "FeUdal Networks for Hierarchical Reinforcement Learning", ARXIV.ORG, 3 March 2017 (2017-03-03), XP080754067
Attorney, Agent or Firm:
RYUKA & PARTNERS (JP)
Download PDF: