Emerging nonvolatile memory technologies have been much studied to realize massively interconnected Spiking Neural Networks (SNNs) due to its high density, and energy efficiency. Nevertheless, most of the previous studies on utilizing such technologies have focused on implementing a traditional pair-based Spike Timing Dependent Plasticity (STDP) rule, which turned out to fail to reproduce multiple physiological experimental results. In this paper, we present a neuromorphic circuit for a three-terminal spintronic synapse with a triplet-based STDP rule, which is a higher order of learning mechanism. Simulation results indicate that the proposed learning circuit for the triplet-based STDP rule significantly improves learning capability in mimicking the various biological experimental data. We introduce a normalized mean-square error E value to evaluate the performance of each of the learning circuits quantitatively. The proposed learning circuit achieves E value of 1.77, which is far better than the conventional one that achieves E value of 12.2.
|Title of host publication||2018 25th IEEE International Conference on Electronics Circuits and Systems, ICECS 2018|
|Publisher||Institute of Electrical and Electronics Engineers Inc.|
|Number of pages||4|
|Publication status||Published - 2019 Jan 17|
|Event||25th IEEE International Conference on Electronics Circuits and Systems, ICECS 2018 - Bordeaux, France|
Duration: 2018 Dec 9 → 2018 Dec 12
|Name||2018 25th IEEE International Conference on Electronics Circuits and Systems, ICECS 2018|
|Conference||25th IEEE International Conference on Electronics Circuits and Systems, ICECS 2018|
|Period||18/12/9 → 18/12/12|
Bibliographical noteFunding Information:
ACKNOWLEDGMENT This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP) (No.2017R1A2B2006679).
© 2018 IEEE.
All Science Journal Classification (ASJC) codes
- Electrical and Electronic Engineering