An Efficient, Wavelet Based, Variable Structure Network.
Abstract
Learning systems, specifically neural networks trained using supervised learning, have become widely accepted and in many cases provide good approximations to the given target function. One of the parameters that tends to make a large difference in the output of the network is its size, or number of units that comprise the network. However, there is often no easy way to determine how large a network should be. A way to combat this problem is through a variable structure algorithm.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1998
- Accession Number
- ADA358547
Entities
People
- Chad A. Tilbury
Organizations
- Air Force Institute of Technology