Nonvolatile Memory (NVM) for AI and Neuromorphic Computing
摘要
This paper introduces a nonvolatile memory (NVM) device that generates multiple states to store multiple values, and discusses its applications for the multiply-accumulate (MAC) operation. The operation of the device is explained using a device model, and the application of the device is implemented using a circuit model. The architecture for vector–matrix multiplication (VMM) to implement the MAC operation is also shown. The various controllable states of this device will aid in processing a large amount of data for current data science, artificial intelligence (AI), quantum computing (QC), and other large data computing and handling.