• JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
 
  Bookmark and Share
 
 
Master's Dissertation
DOI
https://doi.org/10.11606/D.55.2018.tde-06032018-161351
Document
Author
Full name
Marco Antonio Alvarez Vega
Institute/School/College
Knowledge Area
Date of Defense
Published
São Carlos, 1999
Supervisor
Committee
Carvalho, André Carlos Ponce de Leon Ferreira de (President)
Camargo, Heloisa de Arruda
Monard, Maria Carolina
Title in Portuguese
Um Estudo Comparativo de Técnicas de Pruning para Redes Neurais Artificiais
Keywords in Portuguese
Não disponível
Abstract in Portuguese
Redes Neurais Artificiais (RNAs) têm proporcionado uma solução eficiente para uma grande variedade de problemas práticos. Infelizmente, a seleção dos parâmetros ideais para o processo de aprendizado, bem como a escolha da topologia adequada, não são tarefas triviais Geralmente, o processo de escolha do número de parâmetros livres é informal, e as redes são treinadas com diferentes topologias e complexidades até que a de melhor desempenho seja encontrada. Este procedimento nem sempre produz redes de tamanho mínimo, o que em muitos casos inviabiliza a implementação. Nesta dissertação é apresentado um estudo comparativo de diversas técnicas de Pruning, as quais têm como objetivo minimizar a complexidade da rede, sem degradar sua capacidade de generalização. Um grande número de experimentos foi realizado, utilizando diversas técnicas previamente selecionadas. Uma análise dos resultados obtidos é também apresentada, indicando o comportamento das técnicas de Pruning em geral, e identificando as de melhor desempenho.
Title in English
Not available
Keywords in English
Not available
Abstract in English
Artificial Neural Networks offer an efficient solution to a wide variety of practical problems. However, choosing an appropriate topology for an Artificial Neural Network is a difficult task. Generally, the process of choosing the number of free parameters is informal and networks are trained with different complexities and topologies until the one with the best performance be selected. This procedure usually does not generate minimal size networks, which can make their implementation unfeasible. In the present work, a comparative study of Pruning techniques is presented. Such techniques are used to improve the complexity of networks, reducing their size without considerably degrading their generalization ability. A large number of experiments was performed with a set of previously selected techniques. The results were analyzed in order to study the behavior of the Pruning techniques in general, as well as identifying those which provide the best performance.
 
WARNING - Viewing this document is conditioned on your acceptance of the following terms of use:
This document is only for private use for research and teaching activities. Reproduction for commercial use is forbidden. This rights cover the whole data about this document as well as its contents. Any uses or copies of this document in whole or in part must include the author's name.
Publishing Date
2018-03-06
 
WARNING: Learn what derived works are clicking here.
All rights of the thesis/dissertation are from the authors
CeTI-SC/STI
Digital Library of Theses and Dissertations of USP. Copyright © 2001-2024. All rights reserved.