Biomorphic Networks for ATR and Higher-Level Processing

Abstract

During the period of this report work progressed on the design and implementation in analog hardware of a parametrically modulated logistic processing element for use in the construction of fast analog Parametrically Coupled Logistic Map Networks. The design is based on a periodically driven spiking neuron circuit. In preceding work we put forth the hypothesis that the basic functional unit in the cortex is not the single neuron but is the neuronal assembly or netlet and that the behavior of a netlet can be mathematically modeled by a parametrically modulated logistic map, a nonlinear iterative mapping on the unit interval. We also presented evidence of the utility of such processing elements in the modeling of cortical networks and introduced a new family of networks which we named Parametrically Coupled Logistic Map Networks (abbreviated PCLMNs) that are increasingly appearing to be key to efficient simulation and study of higher-level functions performed by the cortex which is the seat of all higher level brain functions. The advantages of developing the theoretical and practical foundations for such Corticonic networks should be obvious especially for the development of new generation of machines with brain-like intelligence that goes beyond the capabilities of present day neural and connectionist models.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Apr 20, 1999
Accession Number
ADA362731

Entities

People

  • Nabil H. Farhat

Organizations

  • University of Pennsylvania

Tags

DTIC Thesaurus Topics

  • Analog Computers
  • Computers
  • Construction
  • Differential Equations
  • Digital Computers
  • Dynamics
  • Electrical Engineering
  • Engineering
  • Frequency
  • High Resolution
  • Identification
  • Intervals
  • Military Research
  • Phase Measurement
  • Recognition
  • Simulations
  • Time Intervals

Readers

  • Neural Network Machine Learning.
  • Neuroscience
  • Systems Analysis and Design