Neuromorphic computing with ReRAM and PCM

Daniele IELMINI, Politecnico di Milano

Brain-inspired computing tasks such as machine learning and pattern recognition , are becoming increasingly popular in the era of Big Data and the internet of things (IoT). Mimicking the brain–type data processing requires to overcome the von Neumann architecture of conventional computers, and adopt high density nanoelectronic synapses for the necessary efficiency and low power consumption. Emerging memory devices such as resistive switching memory (ReRAM) and phase change memory (PCM) provide an ideal synaptic technology due to analog tuning, low power consumption, nonvolatile storage, and scalability. This talk will summarize the current state of ReRAM/PCM synapses for neuromorphic computing, showing a recent neural network for spike-timing dependent plasticity (STDP) capable of online unsupervised learning. Simulated operation at various level of complexity (digital, gray scale, color scale) and an experimental demonstration with HfOx ReRAM devices will be presented.

 
Back to TS.VII.A