Accession Number:

ADA270825

Title:

VLSI Implementation of Neuromorphic Learning Networks

Descriptive Note:

Final rept. 1 Aug 90-31 Mar 93,

Corporate Author:

BELL COMMUNICATIONS RESEARCH INC LIVINGSTON NJ

Personal Author(s):

Report Date:

1993-03-31

Pagination or Media Count:

9.0

Abstract:

The researchers have shown how to rigorously derive deterministic systems from stochastic ones in the Boltzmann machine framework that they are using for their implementations. They have further shown how to search for new learning algorithms suitable for VLSI implementation using a genetic algorithm approach. They-have analyzed the effect of precision constraints such as is found in hardware implementations on the learning and generalization abilities of neural networks. They have studied the learning behavior of neural networks under conditions where they where they can or cannot classify perfectly.

Subject Categories:

  • Electrical and Electronic Equipment
  • Cybernetics

Distribution Statement:

APPROVED FOR PUBLIC RELEASE