Plasticity in MTJ-based synapses
Résumé
Spiking Neural Networks (SNNs) are being widely studied to mimic the intelligence of biological brains. In hardware level, multiple emerging non-volatile memories (NVM) are proposed to model the neural synapses, one of them being Spin-transfer Torque Magnetic Random-Access Memory (STT-MRAM). This device can be used as a building block to build multilevel conducting synapses. In this paper we describe the simulation results of a compound magnetic synapse in nominal situation, and then present the effect of process variability on its performance.
Fichier principal
TIMA conference21 - Plasticity in MTJ-based synapses .pdf (390.57 Ko)
Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)