2019 Fiscal Year Final Research Report
Pulse driven hardware self-organizing map that mimics biological brain
Project/Area Number |
16K00344
|
Research Category |
Grant-in-Aid for Scientific Research (C)
|
Allocation Type | Multi-year Fund |
Section | 一般 |
Research Field |
Soft computing
|
Research Institution | Kansai University |
Principal Investigator |
Hikawa Hiroomi 関西大学, システム理工学部, 教授 (10244154)
|
Project Period (FY) |
2016-04-01 – 2020-03-31
|
Keywords | 自己組織化マップ / ハードゥエア / パルス信号 / 機械学習 / ベクトル量子化 / ニューラルネットワーク |
Outline of Final Research Achievements |
In this research, a self-organizing map (SOM) that mimics the brain was studied. In the biological brain, information is transmitted by exchanging pulse signals between neurons. It also has the property of trying to acquire various knowledge owing to its curiosity. Focusing on the above two points, hardware SOM that imitates the brain, and a new learning algorithm were developed. In order to provide the pulse-based operation, the SOM-dedicated hardware (electronic circuit) that transmits information between neurons by pulse signals and performs necessary calculations with pulse signals, was developed. Regarding the latter, the learning algorithm was modified for the SOM to perform powerful learning when it encounters unfamiliar information and to actively acquire new information.
|
Free Research Field |
情報工学
|
Academic Significance and Societal Importance of the Research Achievements |
自己組織化マップは,画像圧縮や大量のデータの分析などに使われるニューラルネットワークの一種である.専用ハードウェアを開発することで,高性能なコンピュータを使わずに高速なデータ処理が可能となる.また,パルス駆動のハードウェア開発を通して,従来通信関係の回路で使われてきた回路(周波数同期ループ)をニューロンの演算に使用するという,学術的に意義のある研究となった.そして,人間の脳を模倣することでより高性能な処理ができるようになった.今後,脳を模倣する SOM の動作を観察することで人間の脳の働きを解明するツールとしての可能性がある.
|