Committee machines—a universal method to deal with non-idealities in memristor-based neural networks

D. Joksas (), P. Freitas, Z. Chai, W. H. Ng, M. Buckwell, C. Li, W. D. Zhang, Q. Xia, A. J. Kenyon and A. Mehonic ()
Additional contact information
D. Joksas: University College London, Roberts Building, Torrington Place
P. Freitas: Liverpool John Moores University, Liverpool, James Parsons Building, Byrom Street
Z. Chai: Liverpool John Moores University, Liverpool, James Parsons Building, Byrom Street
W. H. Ng: University College London, Roberts Building, Torrington Place
M. Buckwell: University College London, Roberts Building, Torrington Place
C. Li: University of Massachusetts Amherst, 100 Natural Resources Road
W. D. Zhang: Liverpool John Moores University, Liverpool, James Parsons Building, Byrom Street
Q. Xia: University of Massachusetts Amherst, 100 Natural Resources Road
A. J. Kenyon: University College London, Roberts Building, Torrington Place
A. Mehonic: University College London, Roberts Building, Torrington Place

Nature Communications, 2020, vol. 11, issue 1, 1-10

Abstract: Abstract Artificial neural networks are notoriously power- and time-consuming when implemented on conventional von Neumann computing systems. Consequently, recent years have seen an emergence of research in machine learning hardware that strives to bring memory and computing closer together. A popular approach is to realise artificial neural networks in hardware by implementing their synaptic weights using memristive devices. However, various device- and system-level non-idealities usually prevent these physical implementations from achieving high inference accuracy. We suggest applying a well-known concept in computer science—committee machines—in the context of memristor-based neural networks. Using simulations and experimental data from three different types of memristive devices, we show that committee machines employing ensemble averaging can successfully increase inference accuracy in physically implemented neural networks that suffer from faulty devices, device-to-device variability, random telegraph noise and line resistance. Importantly, we demonstrate that the accuracy can be improved even without increasing the total number of memristors.

Date: 2020
References: Add references at CitEc
Citations: View citations in EconPapers (2)

Downloads: (external link)
https://www.nature.com/articles/s41467-020-18098-0 Abstract (text/html)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:nat:natcom:v:11:y:2020:i:1:d:10.1038_s41467-020-18098-0

Ordering information: This journal article can be ordered from
https://www.nature.com/ncomms/

DOI: 10.1038/s41467-020-18098-0

Access Statistics for this article

Nature Communications is currently edited by Nathalie Le Bot, Enda Bergin and Fiona Gillespie

More articles in Nature Communications from Nature
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().