Creation of Self-organizing Cognitive Architecture Integrating Probabilistic Generative Model and Deep Learning

2021 Fiscal Year Final Research Report

• Project/Area Number: 18H03308
• Research Category: Grant-in-Aid for Scientific Research (B)
• Allocation Type: Single-year Grants
• Section: 一般
• Review Section: Basic Section 61040:Soft computing-related
• Research Institution: Ritsumeikan University
• Principal Investigator: Taniguchi Tadahiro 立命館大学, 情報理工学部, 教授 (80512251)
• Co-Investigator(Kenkyū-buntansha): 萩原 良信 立命館大学, 情報理工学部, 講師 (20609416)
• Project Period (FY): 2018-04-01 – 2021-03-31
• Keywords: 機械学習 / 深層学習 / 確率的生成モデル / 認知アーキテクチャ / 自己組織化

Outline of Final Research Achievements

We have proposed a novel framework called Neuro-SERKET as a distributed development framework to develop a self-organizing cognitive architecture that can utilize the advantages of both probabilistic generative models and deep learning. The Neuro-SERKET is an extension of the previously proposed framework, SERKET. This enables us to integrate (deep) probabilistic generative models that have been developed in a distributed manner and to reason about them as a whole. We also have studied the fusion of SLAM and GAN, and the fusion of deep generative models for speech conversion and probabilistic generative models for speech recognition. Through these studies, we built the foundation of a self-organizing cognitive architecture.

Free Research Field

創発システム

Academic Significance and Societal Importance of the Research Achievements

実世界で学習し続けるロボットを生み出すためには視覚や触覚，音声といったマルチモーダル情報を得続け，また行動を意思決定し続ける認知アーキテクチャが必要である．本研究では認知システムの内部で情報が自己組織化するように「学習し続ける」認知アーキテクチャのための基礎理論として，確率的生成モデルと深層学習を融合させるフレームワークの開発を行った．これは機械学習と認知ロボティクスを架橋する研究であるとともに，実世界で活動し続ける自律的な人工知能を生み出すための基盤技術となる．