Analog In-Memory Computing For Deep Learning Inference (Joint MTL/AI HW Seminar)

MTL Seminar Series
Abu Sebastian, IBM Zurich

Abstract

First, I will provide an overview of analog in-memory computing (AIMC) in the context of deep learning inference, emphasizing essential concepts and related terminology. Following that, I will offer a detailed overview of a mixed-signal AIMC prototype chip that utilizes phase-change memory technology and is manufactured using 14nm CMOS technology. Lastly, I will offer a glimpse into the future by highlighting ongoing research efforts, spanning from memory device advancements to novel applications.