Biological Investigations of Adaptive Networks: Neuronal Control of Conditioned Responses
Abstract
Investigations of adaptive neutral networks were conducted using the classically conditioned nictitating membrane response (NMR) of rabbit. Work involved both neurobiological and theoretical approaches based on mathematical models and computer simulation. Recordings were done from single brain stem neurons in awake, behaving animals for the purpose of determining the loci and activity relate to CRs. Computational tools for applying systems analysis to neurophysiological data obtained from single-unit recordings from awake behaving animals were developed. The relationship between single neurons' dynamic behavior and the CR in terms of differential equations and sophisticated correlational analyses based on Fourier and Laplace transform methods was characterized. Theoretical studies revolved around two mathematical models of learning. The Sutton-Barto-Desmond (SBD) model was designed to describe real-time features of the NM CR. A cerebellar network implementation of this model was constructed by combining parametric constraints of the model dictated by behavioral data with constraints based on anatomy and physiology of the cerebellum. The second major theoretical development was the construction of a two-element neural-network architecture that elegantly describes adaptive timing as manifested in the fine-grain temporal characteristics of CRs.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1989
- Accession Number
- ADA211043
Entities
People
- John W. Moore
Organizations
- University of Massachusetts Amherst