Associative memory model that extracts correlated structure of observed images and stores the individual images as continuous attractor

2015 Fiscal Year Final Research Report

• Project/Area Number: 25330298
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Soft computing
• Research Institution: Oita National College of Technology
• Principal Investigator: KIMOTO Tomoyuki 大分工業高等専門学校, 電気電子工学科, 教授 (30259973)
• Co-Investigator(Kenkyū-buntansha): UEZU TATSUYA 奈良女子大学, 大学院人間文化研究科, 教授 (10160160)
• Project Period (FY): 2013-04-01 – 2016-03-31
• Keywords: ホップフィールドモデル / 点アトラクタ / 連続アトラクタ / XYスピン / 位相振動子

Outline of Final Research Achievements

When we observe the faces of a person from different directions, a plurality of images are obtained, but storing these images separately is not a natural information representation. In this study, we were succeeded in storing a plurality of images as a continuous attractor, by improving the Hopfield model. The Hopfield model is one of the memory model of the brain, and is composed of the elements called the Ising spin. We found that continuous attractor occurs by extending the Ising spin to XY spin in the Hopfield model. Furthermore, even if we extend the Ising spin to the phase oscillator, it was found that the continuous attractor occurs.

Free Research Field

ニューラルネットワーク