Spatial-temporal memory effects in nanoelectronic neuromorphic networks

Abstract

The main objective of this project is to carry out a coordinated theoretical and experimental study of spatial-temporal memory (STM) effects in neuromorphic networks based on memristive crossbars. Previous studies indicate that such networks (ÒCrossNetsÓ), performing analog processing of information, may eventually far overcome both their biological prototypes (cortical circuits) and digital artificial neural networks in both density and speed, at manageable power dissipation. Our project will address the key challenges on the way toward practical, large-scale, high-performance CrossNet spatial-temporal memories, in the following areas: 1. General STM theory. We are going to use both approximate analytical approaches and numerical modeling to establish the quantitative relation between the capacity, fidelity, and noise immunity of spiking, CrossNet-based spatial-temporal memories, with two types of pattern recording: ex-situ (using a software precursor) and in-situ (using synaptic spiking-timedependent plasticity Ð STDP). 2. Memristor-specific theory. Based on experimentally characterized properties of the new generation of crossbar-integrated bi-oxide (Al2O3/TiO2-x) memristive devices being fabricated by our Santa Barbara group, we will determine optimal ways to implement the STM in CrossNets using such crossbars. 3. Memristive CrossBar experiments. The results of those analyses will be experimentally verified using relatively small crossbar STM prototypes. Based on the experimental results, we may suggest different, more promising STM implementations.

Document Details

Document Type
DoD Grant Award
Publication Date
Dec 04, 2018
Source ID
W911NF1610302

Entities

People

  • Dmitri Strukov

Organizations

  • Army Contracting Command
  • United States Army
  • University of California, Santa Barbara

Tags

Fields of Study

  • Physics

Readers

  • Computational Modeling and Simulation
  • Electrochemical Surface Science
  • Integrated Circuit Design and Technology.

Technology Areas

  • AI & ML