Neuromorphic Computing with Memristors: From Devices to Integrated Systems

Abstract:

Neuromorphic computing is a concept to use electronic analog circuits to mimic neuro-biological architectures present in the nervous system, which can potentially offer orders of magnitude better power efficiency compared with conventional digital systems. In particular, memristors and memristor crossbar arrays have been widely studied for neuromorphic applications. This thesis work explores the device characteristics and internal dynamics of WOx-based memristor devices, as well as the crossbar array structure and directly integrated hybrid memristor/mixed-signal CMOS circuits for neuromorphic computing.

Neuromorphic applications have been experimentally demonstrated: First, a sparse coding algorithm is implemented in a 32—32 memristor crossbar array, by performing vector-matrix multiplication in physics. Natural image processing and online dictionary learning results are shown. We further fabricated a 54—108 passive memristor crossbar array directly integrated with a custom-designed CMOS chip. With the fully-integrated, reprogrammable chip, we demonstrated multiple models such as perceptron learning, principal component analysis, and also sparse coding, all in one single chip. Besides that, the internal device dynamics, including the short-term memory effect caused by spontaneous oxygen vacancy diffusion, additionally allows us to implement a reservoir computing system to process temporal information. Our effort to solve other challenging problems with memristor-based system will also be discussed.