Temporal Processing with Neural Networks.
Abstract
The research carried out under this contract focussed on four efforts, all involving the processing of temporal sequences by neural networks (1-3) or the effect of imposing a spatio-temporal gradient on network learning (4): (1) Assessing alternative neural network techniques for problems involving temporal coding. (2) Development of tools for analysing recurrent networks, so that the solutions of successfully trained networks can be better understood; (3) Development of a dynamical systems theory approach to computation in recurrent networks. (4) Development of biologically and cognitively plausible techniques for enhancing training.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 25, 1998
- Accession Number
- ADA347549
Entities
People
- Jeffrey L. Elman
Organizations
- University of California, San Diego