Task 005 - Neural Fabrics

ABSTRACT:

Resistive RAM (RRAM) based in-memory computing (IMC) has emerged as a promising paradigm for efficient deep neural network (DNN) acceleration. However, the multi-bit RRAMs often suffer from non-ideal characteristics such as drift and retention failure against temperature changes, leading to significant inference accuracy degradation. In this work, we present a new temperature-resilient RRAM-based IMC scheme for reliable DNN inference hardware. From a 90nm RRAM prototype chip, we first measured the retention characteristics of multi-level HfO2 RRAMs at various temperatures up to 120°C, then rigorously modeled the temperature-dependent RRAM retention behavior. We propose a novel and efficient DNN training/inference scheme along with the system-level hardware design to resolve the temperature-dependent retention issues with one-time DNN deployment. Employing the proposed scheme on 256 x 256 RRAM array with the circuit-level benchmark simulator NeuroSim, we demonstrate robust RRAM IMC based DNN inference where >30 % CIFAR-10 accuracy and >60 % TinyImageNet accuracy are recovered against temperature variations.

BIO:

Jian Meng received the B.S degree from Portland State University, Portland, USA, in 2019. He is currently pursuing the Ph.D. degree with the School of Electrical, Computer and Energy Engineering, Arizona State University, Tempe, AZ, USA. His current research focuses on the deep neural network compression optimization, hardware – software co-design with emerging non-volatile memory technologies, and event-based object detections.