Memristive Optimizer for the Assignment Task

Abstract

This work shows an analog CMOS matrix compatible with memristors, where they work as dynamic resistors. They are also programmed to initial state values according to one assignment optimization task, taken here as an application vehicle. The whole matrix accomplishes a parallel competitive computation, where a winner-take-all mechanism is inherent by architecture. The expected complexity due to area in silicon of this memristive circuit and its connectivity are moderate, as is evident from observing its electrical diagram. Also, its performance is acceptable as proven via SPICE simulations, using 0.5-micron CMOS process. Lower processing time might be possible with latest CMOS technologies.