Multinary Data Processing Based on Nonlinear Synaptic Devices

Abstract

Stasis weight (G s) that is an independent weight on applied input pulse amplitude was demonstrated for multinary data processing in a synaptic device-based neuromorphic system. Because multinary data is implemented as various input pulse amplitudes, the G s is necessary for high inference accuracy. Typically, synaptic devices exhibit nonlinear current?voltage characteristics (nonlinear device) and have different weight values depending on the applied input pulse amplitudes (G ns). Therefore, to achieve high inference accuracy, we proposed pulse modulation circuits that can transform the pulse amplitude into pulse width or number. As a result, the G s was obtained from the nonlinear device possessing the G ns, and the inference accuracy of the simulated MNIST data set was obviously improved from 29.34% to 97.6%. ? 2021, The Minerals, Metals & Materials Society.