Memristive Neuromorphic Computing Elements

In this work, fundamental components of a neuromorphic system, including nanoelectronic equivalents of synapses and neurons with signal (spike) integration and generation functionalities, were demonstrated using oxide memristor-based devices. Both non-volatile and volatile memristors were demonstrated, with control over oxide stoichiometry, which corresponded to the electrical behavior. Circuits comprised of these devices were simulated and shown to exhibit behaviors analogous to those of biological spiking neurons. These devices and circuits form the basis of systems that can perform specialized computing tasks far more efficiently than conventional Si CMOS-based systems in terms of SWaP (size, weight and power) criteria.