Name
Abstract | ||
|---|---|---|
OxRRAM is being considered as synapse in neuromorphic systems in multiple ways. For analog neural network weight storage, OxRRAM resistance variability poses a significant challenge at low currents. However, alternative ‘cortical’ learning algorithms can tolerate or even exploit the stochastic behavior of the device at low power. This paper aims at providing an accurate kinetic description of the low voltage transients in OxRRAM. To model the relevant stochastic effects, we extend the hourglass model with a power-dependent filament shuffling rate and a normally distributed activation energy. Including these elements improves the original hourglass model and allows for resistance distribution simulations at low voltage as well as reproducing resistance-time transient RESET traces with inclusion of the intrinsic stochastic variability. |
Year | DOI | Venue |
|---|---|---|
2019 | 10.1109/IRPS.2019.8720555 | 2019 IEEE International Reliability Physics Symposium (IRPS) |
Keywords | Field | DocType |
component,OxRRAM,low voltage gradual RESET,neuromorphic,stochastic learning,Hourglass model | Neuromorphic engineering,Electronic engineering,Low voltage,Engineering,Kinetic energy | Conference |
ISSN | ISBN | Citations |
1541-7026 | 978-1-5386-9504-3 | 0 |
PageRank | References | Authors |
0.34 | 0 | 7 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| J. Doevenspeck | 1 | 0 | 0.68 |
| Robin Degraeve | 2 | 7 | 3.20 |
| A. Fantini | 3 | 0 | 0.68 |
| Peter Debacker | 4 | 32 | 9.04 |
| D. Verkest | 5 | 372 | 37.99 |
| R. Lauwereins | 6 | 180 | 19.03 |
| Wim Dehaene | 7 | 874 | 116.42 |