Energy Efficient Neuromorphic Algorithm Training with Analog Memory Arrays

Abstract

Deep learning is empowering new frontiers in autonomous control for government microelectronic systems. Deep networks have unprecedented accuracy on many recognition tasks, but difficult to execute and nearly impossible to train or update deep networks on SWaP constrained systems operating in the field. Analog in-memory training offers a promising route for unprecedented low SWaP deep network training and execution for embedded systems with excellent performance in radiation environments.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Mar 25, 2019
Accession Number
AD1075392

Entities

People

  • Alec A. Yalin
  • Alex H. Hsia
  • Chris Bennett
  • Conrad D. James
  • David R. Haghart
  • Edward Bielejec
  • Elliot J. Fuller
  • George Vizkelethy
  • Hugh Barnaby
  • J. L. Taggart
  • Matthew Marinella
  • Robin B. Jacobs-gedrim
  • Sapan Agarwal

Organizations

  • Sandia National Laboratories

Tags

Communities of Interest

  • Advanced Electronics
  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Accuracy
  • Algorithms
  • Computations
  • Deep Learning
  • Energy Efficiency
  • Image Recognition
  • Ions
  • Learning
  • Memory Devices
  • Neural Networks
  • Physical Properties
  • Radiation
  • Resistance
  • Semiconductors
  • Standards
  • Switching
  • Terminals

Fields of Study

  • Computer science

Readers

  • Distributed Systems and Data Platform Development
  • Integrated Circuit Design and Technology.
  • Team-Based Human-Centered Cognitive Task Decision Making and Information Performance.

Technology Areas

  • AI & ML
  • AI & ML - Autonomous Systems
  • AI & ML - Neural Networks
  • Microelectronics