2020 Volume 89 Issue 1 Pages 41-45
A Resistive Analog Neuromorphic Device (RAND) is a device that can store electrical resistance values in a non-volatile manner by utilizing the resistive switch effect in insulating oxides. By using this RAND, it is possible to realize small-sized and low-power consumption brain-inspired information processing. This article describes the overview of RAND, the mechanism of the analog resistance change, and an architecture using RAND. An example of the RAND circuit performance evaluation is also shown. We hope you will feel the necessity and inevitability of developing software and hardware synchronously for AI research.