On the fault tolerance of a clustered single-electron neural network for differential enhancement

• Oya Takahide

  Graduate School of Information Science and Technology, Hokkaido University

• Schmid Alexandre

  Microelectronic Systems Laboratory, Swiss Federal Institute of Technology (EPFL)

• Asai Tetsuya

  Graduate School of Information Science and Technology, Hokkaido University

• Leblebici Yusuf

  Microelectronic Systems Laboratory, Swiss Federal Institute of Technology (EPFL)

• Amemiya Yoshihito

  Graduate School of Information Science and Technology, Hokkaido University

Abstract

A clustered neural network, in which neuronal information is represented by a cluster (population of neurons), rather than a single neuron, is a possible solution to construct fault-tolerant single-electron circuits. We designed single-electron circuits based on a clustered neural network that performs differential enhancement where differences between the cluster's outputs receiving various magnitudes of inputs are enhanced after the processing. Simulation results showed that the degradation of the performance of the clustered single-electron neural network was significantly lower than that of a non-clustered network, which indicates that this approach is one possible way to construct fault-tolerant computing systems on nanodevices.

Journal

• IEICE Electronics Express

  IEICE Electronics Express 2 (3), 76-80, 2005

  The Institute of Electronics, Information and Communication Engineers

Keywords

• single-electron circuit
• neural network
• fault tolerance

Details

• CRID: 1390282680189395456
• NII Article ID: 130000088203
• DOI: 10.1587/elex.2.76
• ISSN: 13492543
• Web Site: http://www.jstage.jst.go.jp/article/elex/2/3/2_3_76/_pdf
• Text Lang: en
• Data Source:
  • JaLC
  • Crossref
  • CiNii Articles
• Abstract License Flag: Disallowed