We present a new electronic synapse for neuromorphic computing consisting of a 1T1R structure based on HfO2RRAM technology, and capable of STDP and pattern learning. Power consumption is reduced by adopting short POST spike and burst-mode integration. MNIST classification shows promising learning and classification efficiency. These results support RRAM as an enabling technology for low-power neuromorphic hardware.
Novel RRAM-enabled 1T1R synapse capable of low-power STDP via burst-mode communication and real-Time unsupervised machine learning
Ambrogio, S.;Balatti, S.;Milo, V.;Carboni, R.;Wang, Z.;Ielmini, D.
2016-01-01
Abstract
We present a new electronic synapse for neuromorphic computing consisting of a 1T1R structure based on HfO2RRAM technology, and capable of STDP and pattern learning. Power consumption is reduced by adopting short POST spike and burst-mode integration. MNIST classification shows promising learning and classification efficiency. These results support RRAM as an enabling technology for low-power neuromorphic hardware.File in questo prodotto:
| File | Dimensione | Formato | |
|---|---|---|---|
|
2016_vlsi.pdf
Accesso riservato
:
Publisher’s version
Dimensione
1.29 MB
Formato
Adobe PDF
|
1.29 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
