Novel Information Processing Architectures for Coarse-Grained Devices

2017 Fiscal Year Final Research Report

• Project Area: Molecular Architectonics: Orchestration of Single Molecules for Novel Function
• Project/Area Number: 25110015
• Research Category: Grant-in-Aid for Scientific Research on Innovative Areas (Research in a proposed research area)
• Allocation Type: Single-year Grants
• Review Section: Science and Engineering
• Research Institution: Hokkaido University
• Principal Investigator: Asai Tetsuya 北海道大学, 情報科学研究科, 教授 (00312380)
• Co-Investigator(Kenkyū-buntansha): 大矢 剛嗣 横浜国立大学, 大学院工学研究院, 准教授 (30432066)
• Project Period (FY): 2013-06-28 – 2018-03-31
• Keywords: 粗粒素子 / ゆらぎ / 分子ネットワーク / リザーバ計算 / コンシステンシ / セルオートマトン

Outline of Final Research Achievements

Novel computing methods and architectures were constructed for coarse-grained molecular devices and materials that utilize noise and fluctuations. The following two computing methods were developed. A cellular-automaton model that imitates neuromorphic spike generation in a molecular spiking neural networks was developed. By using dynamics of the model, performance of 'reservoir computing' was evaluated by using standard benchmarks for memorizing complex temporal sequences. The results showed that the model was able to learn complex temporal sequences at high precision. For the physical demonstration, one had to resolve two issues, i.e., reservoir's initial-value dependence and difficulty in reproduction of complex temporal sequences. To resolve this problem, a nonlinear phenomena, called consistency, was introduced in the reservoir, which resulted in successful generation of initial-value independent and reproductive complex temporal sequences.

Free Research Field

集積回路工学