Non-Linear Analysis of Visual Cortical Neurons.
Abstract
Quantitative procedures were developed for testing block-structured models for multi-input nonlinear visual circuits studied with spatiotemporal white noise. A linear-nonlinear (LN) model test index was found to be suitable for classifying cells as simple versus complex. Although simple cells were better modeled as LN systems than complex cells, most simple cells deviated considerably from LN behavior. A nonlinearity of cortical origin would appear to be responsible, possibly activated more strongly by broadband noise than by sinewave grating stimuli. Also, two classes of binocular complex cells were identified. Whereas all binocular complex cells necessarily have a non-zero second-order same-eye interaction kernel, their second-order cross-eye interaction kernel could, it was found, be either non-zero or identically zero. Binocular vision, nonlinear system identification, neural network.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 07, 1992
- Accession Number
- ADA250233
Entities
People
- Daniel A. Pollen
- James P. Gaska
- Lowell D. Jacobsen
Organizations
- University of Massachusetts Medical School