Abstract
Stochastic computing is an old and unconventional computing technique that is finding promising new applications in image processing and the handling of complex error-correcting codes. Stochastic circuits offer an alternative to conventional digital circuits because of their extremely small size and inherent noise tolerance. They are also well-suited to meeting the requirements of emerging nanoscale technologies where non-deterministic behavior due to manufacturing defects and soft errors cannot be ignored. Error analysis of stochastic circuits, however, has received little attention and remains a largely open problem, especially when multiple errors affecting both the data sources and the stochastic circuits can occur in the course of a computation. This paper attempts to analyze stochastic circuits under various error conditions, and to compare their behavior to that of conventional circuits under similar error conditions. We use probabilistic transfer matrices for this analysis, complemented by circuit simulation. Our results indicate that stochastic circuits provide significantly better error tolerance under severe error conditions.
About the authors
Te-Hsuan Chen is a Ph. D. candidate in Computer Science and Engineering at the University of Michigan. He received his B. S. and M. S. degrees in electrical engineering from the National Tsing Hua University, Hsinchu, Taiwan. His research interests include computer-aided design, design and testing of VLSI circuits, error-correcting coding and unconventional computing methods.
2260 Hayward Street, University of Michigan, Ann Arbor, MI 48109, USA,Tel.: +1 734 763-6411
Armin Alaghi received his B. Sc. degree in electrical engineering (2006) and M. Sc. degree in computer architecture (2009) from the University of Tehran, Iran. He is currently a Ph. D. candidate at the University of Michigan. His research interests include digital system design and testing, approximate computing, and embedded computing. His current research focuses on stochastic computing.
2260 Hayward Street, University ofMichigan, Ann Arbor, MI 48109, USA,Tel.: +1 734 763-6411
Prof. John P. Hayes received the B. E. degree from the National University of Ireland, Dublin, and the M. S. and Ph. D. degrees from the University of Illinois, Urbana-Champaign, all in electrical engineering. He is currently Professor of EECS and holder of the Claude E. Shannon Chair of Engineering Science at the University of Michigan. His teaching and research interests include computer-aided design, verification and testing of VLSI circuits, fault-tolerant computer architecture, and unconventional computing methods.
2260 Hayward Street, University ofMichigan, Ann Arbor, MI 48109, USA,Tel.: +1 734 763-0386,Fax: +1 734 763-4617
©2014 Walter de Gruyter Berlin/Boston
