Memristive Crossbar Arrays for Analog In-Memory Computing and Neuromorphic Engineering

MTL Seminar Series
Qiangfei Xia (UMass Amherst)

Abstract

Memristive crossbar arrays perform multiply–accumulate operations by utilizing physical laws, boosting power efficiency and computing throughput. Based on ionic motion, memristive devices are also bio-realistic emulators for artificial neurons and synapses. In this talk, I will first introduce a computing memristor for analog in-memory computing. I will then showcase an AI accelerator made by integrating memristive crossbars with CMOS circuitry. Finally, I will discuss a diffusive memristor and its application in neuromorphic engineering.