Analog Neural Computing With Super-Resolution Memristor Crossbars

Memristor crossbar arrays are used in a wide range of in-memory and\nneuromorphic computing applications. However, memristor devices suffer from\nnon-idealities that result in the variability of conductive states, making\nprogramming them to a desired analog conductance value extremely difficult as\nthe device ages. In theory, memristors can be a nonlinear programmable analog\nresistor with memory properties that can take infinite resistive states. In\npractice, such memristors are hard to make, and in a crossbar, it is confined\nto a limited set of stable conductance values. The number of conductance levels\navailable for a node in the crossbar is defined as the crossbar's resolution.\nThis paper presents a technique to improve the resolution by building a\nsuper-resolution memristor crossbar with nodes having multiple memristors to\ngenerate r-simplicial sequence of unique conductance values. The wider the\nrange and number of conductance values, the higher the crossbar's resolution.\nThis is particularly useful in building analog neural network (ANN) layers,\nwhich are proven to be one of the go-to approaches for forming a neural network\nlayer in implementing neuromorphic computations.\n