Resistance-Based Embedded Non-Volatile Memories as Synaptic Devices in Spiking Neural Networks

This paper presents prospect and challenges of using state-of-the-art memory technologies in neuro-inspired computing. As it has become critical to move beyond the Von-Neumann architecture, emulation of biological system has been sought after. In this paper, characteristics of a desired synaptic device required for constructing neural network are examined to highlight what functionalities of synapses can be emulated with new non-volatile memory (NVM) devices, what the basic requirements are to realize artificial inorganic neurons and synapses, and which material and device structures can be used for this purpose.