Sep 16 (Thu) @ 3:00pm: "Energy-Efficient Neuromorphic Computing with CMOS-Integrated Memristive Crossbars," Zahra Fahimi, ECE PhD Defense

Abstract

In-memory computing is a promising solution for implementing energy-efficient neuromorphic systems since it minimizes data transportation between memory and the processing units. The major component in developing neuromorphic circuits is a nanoscale memory device, which is responsible for weight storage and analog computation. Resistive Random-Access Memory (RRAM) is one of the most promising memory candidates due to its long-term retention, analog storage, low-power operation, and compact nanoscale footprint.

In this research, we explore the nonidealities of RRAM technology, such as temperature dependency, noise, stuck-at-fault, and tunning error, and their impact on the accuracy of neuromorphic hardware implementation. We show that these imperfections may significantly degrade the inference accuracy of neuromorphic circuits. To mitigate them, we have proposed a holistic approach based on hardware-aware training in which modifications are done in tunning, circuit, and training phase (ex-situ) of hardware development. The proposed method significantly decreases the accuracy drop across the 25–100 °C temperature range, allows 2.5× to 9× improvement in energy consumption of the memory arrays during inference, and improves the defect tolerance by >100×.

We also study the impact of device uniformity in passive memristive circuits and the tradeoffs between computing accuracy, crossbar size, switching threshold variations, and target precision. Nonidealities are investigated in two representative deep neural networks, and several solutions are proposed to enhance the performance. These techniques allow us to implement advanced deep neural networks (DNNs) with almost no accuracy drop, using state-of-the-art analog 0T1R technology.

In the last part of this research, we focus on integrating passive and active RRAM with CMOS circuits for implementing efficient demos for various applications such as neural networks. First, focusing on passive technology, we show the building block circuit that facilitates the various operation modes of RRAM circuits. We discuss several neuromorphic networks and prototype demos with integrated analog passive RRAM and CMOS. The designs are fabricated in two wafer-scale tapeout runs in 180 nm CMOS technology, and preliminary encouraging experimental results are obtained. Finally, we demonstrate a massive DNN accelerator fabricated in a standard 65 nm CMOS process with integrated active analog RRAM devices.

Bio

Zahra Fahimi received her M.Sc. degree in Electronic Engineering from Isfahan University of Technology, Iran. She is currently a Ph.D. candidate in Prof. Strukov’s lab.  

Hosted by: Prof. Dmitri Strukov

Submitted by: Zahra Fahimi <zfahimi@ucsb.edu>