Full CMOS-Memristor Implementation of a Dynamic Neuromorphic Architecture

Abstract

Neuromorphic computing systems seek to emulate biological neural functionality emulated in either software or electrical hardware. A key function for such systems is their ability to learn and adapt. In the human brain, such learning and adaptation is achieved via modulation of synaptic connections between different neurons. Memristors (implemented as resistive random access memory or ReRAM) have great potential to provide synaptic functionality for neuromorphic chip architectures. Under an AFRL-sponsored program, we have developed a unique memristor-CMOS hybrid system for implementing a dynamic adaptive neural network array, also known as mrDANNA. Most recently, our effort has moved from a software-based emulator, to FPGA implementation, and finally to the design, tapeout, and fabrication of this unique, adaptive approach to neuromorphic computing.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Mar 12, 2018
Accession Number
AD1052240

Entities

People

  • Gangotree Chakma
  • Garrett S. Rose
  • J. Murray
  • James S. Plank
  • Joseph Van Nostrand
  • Karsten Beckmann
  • Mark Dean
  • Mussabir Adnan
  • Nathaniel C Cady
  • Ryan Weiss
  • Sagarvarma Sayyaparaju
  • Sherif Amer
  • Wilkie Olin-Ammentorp

Tags

Communities of Interest

  • Advanced Electronics

DTIC Thesaurus Topics

  • Air Force
  • Air Force Research Laboratories
  • Application Software
  • Artificial Intelligence Software
  • Circuits
  • Computer Science
  • Computers
  • Data Storage Systems
  • Electrical Engineering
  • Engineering
  • Memristors
  • Military Research
  • Networks
  • Neural Networks
  • Reservoir Computing
  • Reservoirs
  • Supervised Machine Learning

Readers

  • Integrated Circuit Design and Technology.
  • Neural Network Machine Learning.
  • Parallel and Distributed Computing.

Technology Areas

  • AI & ML
  • AI & ML - Neural Networks