On Comparing Neural Net Training Paradigms via Graded Pattern Recognition Tasks

Published In

Neural Networks

Document Type

Citation

Publication Date

1-1-1988

Abstract

A number of training paradigms for neural nets are under investigation by various researchers around the world. Usually, the efforts (as reported in the literature) focus on one of the paradigms, each contributing to the array of results being accumulated. The project to be reported in this paper, on the other hand, focuses on developing comparative information about a number of the paradigms. The training tasks for the networks are based on a set of pattern recognition problems. The data being used was created some 14 years ago while the author was at NASA developing machine implementable pattern recognition algorithms using then current (non neural-network) methodologies. The base data was in the form of aerial photographic imagery, and the task was to classify the images into one of five land use categories.

Locate the Document

https://doi.org/10.1016/0893-6080(88)90072-X

DOI

10.1016/0893-6080(88)90072-X

Persistent Identifier

https://archives.pdx.edu/ds/psu/37314

Share

COinS