公开(公告)号：US20230215133A1

公开(公告)日：2023-07-06

申请号：US17901148

申请日：2022-09-01

Applicant: Korea Advanced Institute of Science and Technology

Inventor： Sang Wan LEE ,  Shin Young JOO

IPC: G06V10/70 ,  G06N3/04

CPC classification number: G06V10/70 ,  G06N3/0454

Abstract: Disclosed is a method for self-supervised reinforcement learning (RL) by analogy executed by a computer device, the method including configuring a self-supervised RL with analogical reasoning (SRAR) model; and learning a policy for problem solving in a situation in which a task domain changes using the configured SRAR model.