Neuromorphic computing with multi-memristive synapses.
- Boybat I IBM Research - Zurich, Säumerstrasse 4, 8803, Rüschlikon, Switzerland. ibo@zurich.ibm.com.
- Le Gallo M IBM Research - Zurich, Säumerstrasse 4, 8803, Rüschlikon, Switzerland.
- Nandakumar SR IBM Research - Zurich, Säumerstrasse 4, 8803, Rüschlikon, Switzerland.
- Moraitis T IBM Research - Zurich, Säumerstrasse 4, 8803, Rüschlikon, Switzerland.
- Parnell T IBM Research - Zurich, Säumerstrasse 4, 8803, Rüschlikon, Switzerland.
- Tuma T IBM Research - Zurich, Säumerstrasse 4, 8803, Rüschlikon, Switzerland.
- Rajendran B Department of Electrical and Computer Engineering, New Jersey Institute of Technology, Newark, NJ, 07102, USA.
- Leblebici Y Microelectronic Systems Laboratory, EPFL, Bldg ELD, Station 11, CH-1015, Lausanne, Switzerland.
- Sebastian A IBM Research - Zurich, Säumerstrasse 4, 8803, Rüschlikon, Switzerland. ase@zurich.ibm.com.
- Eleftheriou E IBM Research - Zurich, Säumerstrasse 4, 8803, Rüschlikon, Switzerland.
- 2018-06-30
Published in:
- Nature communications. - 2018
Action Potentials
Animals
Biomimetic Materials
Electric Conductivity
Electronics
Humans
Models, Neurological
Neural Networks, Computer
Synapses
Unsupervised Machine Learning
English
Neuromorphic computing has emerged as a promising avenue towards building the next generation of intelligent computing systems. It has been proposed that memristive devices, which exhibit history-dependent conductivity modulation, could efficiently represent the synaptic weights in artificial neural networks. However, precise modulation of the device conductance over a wide dynamic range, necessary to maintain high network accuracy, is proving to be challenging. To address this, we present a multi-memristive synaptic architecture with an efficient global counter-based arbitration scheme. We focus on phase change memory devices, develop a comprehensive model and demonstrate via simulations the effectiveness of the concept for both spiking and non-spiking neural networks. Moreover, we present experimental results involving over a million phase change memory devices for unsupervised learning of temporal correlations using a spiking neural network. The work presents a significant step towards the realization of large-scale and energy-efficient neuromorphic computing systems.
- Language
-
- English
- Open access status
- gold
- Identifiers
-
- DOI 10.1038/s41467-018-04933-y
- PMID 29955057
- Persistent URL
- https://folia.unifr.ch/global/documents/16402
Statistics
Document views: 7
File downloads:
- fulltext.pdf: 0