Abstract | ||
---|---|---|
This study proposes a technique for programming a dense memristor crossbar array without isolation transistors (0T1M) in order to achieve ex-situ training of a neural network. Programming memristors to a specific resistance level requires an iterative process needing the reading of individual memristor resistances due to memristor device stochasticity. This paper presents a circuit to read individual resistances from a 0T1M crossbar and a method to map neuron synaptic weights into a novel neural circuit to enable ex-situ training. The results show that we are able to train the resistances in a 0T1M crossbar and that the 0T1M system is about 93% smaller in area than 1T1M systems. |
Year | DOI | Venue |
---|---|---|
2015 | 10.1109/NANOARCH.2015.7180590 | Proceedings of the 2015 IEEE/ACM International Symposium on Nanoscale Architectures (NANOARCH´15) |
Keywords | Field | DocType |
Memristor crossbars,deep neural networks,pattern recognition,low power system | Memristor,Physical neural network,Iterative and incremental development,Computer science,Neuromorphic engineering,Electronic engineering,Memistor,Artificial neural network,Transistor,Crossbar switch | Conference |
ISSN | Citations | PageRank |
2327-8218 | 3 | 0.40 |
References | Authors | |
12 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Raqibul Hasan | 1 | 76 | 8.74 |
Chris Yakopcic | 2 | 140 | 13.10 |
Tarek M. Taha | 3 | 280 | 32.89 |