Energy-Efficient On-Chip Analysis for Radiation Detection Applications Using Neuromorphic Algorithms and Systems

Abstract

Basic research is proposed on energy-efficient neuromorphic algorithms, architectures, and hardware capable of analyzing data generated by spectroscopic sensors with minimal power consumption. During the Base Period, we will develop a new application-specific neuromorphic algorithm inspired by a locally competitive spiking sparse approximation, build small-scale functional prototypes incorporating filamentary resistive random-access memory (ReRAM) arrays as a proof-of-concept, and test them in a real-world setup. During the Option Period, we will develop, build, and test more advanced devices and algorithms that directly harness the device properties, such as reservoir computing, as well as improve and optimize the prototypes. Demonstration hardware will incorporate nonfilamentary ReRAM arrays.

Document Details

Document Type
DoD Grant Award
Publication Date
Jul 16, 2019
Source ID
HDTRA11810009

Entities

People

  • Marek Osinski

Organizations

  • Defense Threat Reduction Agency
  • University of New Mexico

Tags

Readers

  • Integrated Circuit Design and Technology.
  • Neural Network Machine Learning.
  • Software Engineering