A Hybrid Optimisation Strategy for Adaptive Feed-Forward Layered Networks

Abstract

This British document considers an optimisation strategy for solving for the weight values of a general adaptive feed-forward layered network with linear output units. The weight values of the final layer of the network are determined using linear techniques based on a singular value decomposition of the outputs of the hidden units. The set of weight values governing the transformation of the data prior to this final layer is found using a nonlinear optimisation strategy. This memorandum considers various nonlinear optimisation strategies combined with the linear method and compares the performance of this hybrid approach with previous work which solves for all the weights of the network using nonlinear techniques on a range of problems.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Sep 09, 1988
Accession Number
ADA202067

Entities

People

  • Andrew Webb
  • David Lowe

Organizations

  • Royal Signals and Radar Establishment

Tags

Communities of Interest

  • Air Platforms
  • Materials and Manufacturing Processes
  • Sensors

DTIC Thesaurus Topics

  • Algorithms
  • Arrays
  • Automated Speech Recognition
  • Classification
  • Computer Programming
  • Covariance
  • Data Science
  • Decomposition
  • Discriminant Analysis
  • Equations
  • Focal Planes
  • Information Science
  • Numbers
  • Optimization
  • Probability
  • Test Sets
  • Transfer Functions

Readers

  • Linear Algebra
  • Neural Network Machine Learning.
  • Systems Analysis and Design