Emerging memory technologies for neuromorphic hardware

Abstract This chapter deals with the role that emerging nonvolatile resistive memory technologies (ReRAM) play in the implementation of optimized neuromorphic hardware as a highly promising solution for future ultralow-power cognitive systems. The chapter is organized as follows. We start with an introduction about status and main challenges of current artificial intelligent systems. The key reasons to distribute intelligence over the whole communication network are then discussed, underlining the need for low-power solutions based on specialized embedded hardware. We then show that emerging technologies (and in particular novel resistive memories), coupled with new brain-inspired paradigms, such as spike coding and spike-time-dependent plasticity, have extraordinary potential to provide intelligent features in hardware, approaching the way knowledge is created and processed in the human brain. After a brief survey of the main features and trends of resistive memories (in particular phase-change-, oxide-resistive-, and conductive-bridge-memories), we will introduce the advantages of using ReRAM as synapses to emulate plasticity in spiking neural networks. Several applications (as visual-pattern extraction, bio-signal sorting, etc.) are discussed. Finally, we conclude with our vision of the enabled future directions and the main challenges which should be tackled to exploit the full potential of brain-inspired technologies.