Pulse Coded Biologically Motivated Neural-Type MOS Circuits

Abstract

During this period improvements have been made in the pulse coded neuron which realizes all 16 binary valued logic functions of two inputs. Because the original circuit is inconvenient for VLSI, a study has been made of how to delete the filters used in the previously reported neuron. From this we have been led to a new pulse coded neuron which shows promise for realizing any Hopfield tyoe of neural network in the pulse coded framework. At this point the new pulse coded neuron has been shown to allow us to also generate 14 of the 16 binary valued logic functions without the use of the inconvenient filters of the previous pulse coded neuron. Also using the new neuron a two input MAXNET has been realized. A major problem in implementing Hopfield types of neural networks in the pulse coded philosophy is how to encode general weights. A promising technique to solve this problem is under investigation.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Apr 29, 1991
Accession Number
ADA235433

Entities

Organizations

  • University of Maryland

Tags

Communities of Interest

  • Advanced Electronics

DTIC Thesaurus Topics

  • Amino Acids
  • Arithmetic
  • Arithmetic Units
  • Chemistry
  • Circuits
  • Couplings
  • Electrical Properties
  • Electronic Components
  • Integrated Circuits
  • Membrane Potentials
  • Military Research
  • Networks
  • Neural Networks
  • Neurons
  • Proteins
  • Resistors
  • Theses

Readers

  • Computer Programming and Software Development.
  • Neural Network Machine Learning.
  • Systems Analysis and Design

Technology Areas

  • AI & ML
  • AI & ML - Machine Learning Algorithms
  • AI & ML - Neural Networks