Inaugural Address from the Program Leader
Research Organization of the Program
Multi-Disciplinary COE-student Program
Research Activities and Cross-Disciplinary Units of Sub-Projects
Synopsis of Research Activities
Research Activities of Each Sub-Project
International conferences (BrainIT)
Edited Books
Wrapping Up the 21st Century COE Program
HOME
Research Activities of Each Sub-Project
[ Construction and device implementation of internal models with mnSOM ]
Sub-project leader : Furukawa
Inner model acquisition by using self-organizing architecture with cortex-like network
(1) Abstract
This sub-project team has developed following three self-organizing architectures inspired by the functions and the structures of the cortex. These architectures were applied to autonomous robots to acquire the inner models of outer environment in self-organizing manner.
(1) The modular network SOM [1][2], which has the arrayed network of information processing modules like the column-structure of the cortex.
(2) The higher-rank SOM (SOMn) [3][4], which self-organizes a multifaceted data representation.
(3) The SElf-Evolving Modular network (SEEM), which self-evolves the network based on the mnSOM and the SOMn.
These architectures have been developed by Dr. Furukawa (theory and modeling field) and Dr. Tokunaga (modeling and robotics field) with the cooperation of Dr. Natsume (physiology field). Dr. Tamukoh (device field) implemented SOMn to VLSI. In the application area, Dr. Tokunaga applied SOMn and SEEM to autonomous robots, and the robots succeeded to acquire an inner map of the outer world in self-organizing manner. Dr. Ishii and his colleagues (robotics field) applied the mnSOM to adaptive control of under-water autonomous robots. The mnSOM was also applied to autonomous robots in the Ishikawa’s sub-project.
(2) Members
T. Furukawa (Theory and modeling field)
K. Tokunaga (Modeling and robotics field)
H. Tamukoh (Device field)
K. Natsume (Physiology field)
K. Ishii (Robotics field)
Page top ↑
(3) Figures
Fig.1: The modular network SOM (mnSOM) Fig.2: The self-organizing adaptive controller base on the mnSOM
Movie 1
(MPEG-4:760KB)
: Pole balancing (inverted pendulum) ]
Fig.3: The multifaceted representation of face images organized by SOM2
(Block-diagram) (Result)
Fig.4: The block-diagram of the SOM2 hardware and an organized product manifold
Page top ↑
Fig.5 : The self-evolving modular network (SEEM) Fig.6 and Movie 2: The inner model acquired by an autonomous robot with SEEM.
Movie 2
(AVI:14MB)
References
[1] K. Tokunaga, T. Furukawa, S. Yasui, Modular Network SOM : Self-Organizing Maps in Function Space, Neural Information Processing - Letters and Reviews, Vol.9, No.1, pp.15-22, 2005
[2] K. Tokunaga and T. Furukawa, Modular network SOM: Theory, algorithm and applications, Lecture Notes in Computer Science, Vol.4232, pp.958-967, 2006
[3] T. Furukawa, SOM of SOMs : Self-Organizing Map Which Maps a Group of Self-Organizing Maps, Lecture Notes in Computer Science, Vol.3696, pp.391-396, 2005
[4] Tetsuo Furukawa, Self-Organizing Homotopy Network, Proceedings of the 6th Int. Workshop on Self-Organizing Maps (WSOM 2007), Bielefeld/Germany, 2007
HOME | Back | Page top ↑