Penalty-optimal brain surgeon process and its optimize algorithm based on conjugate gradient

Cuijuan Wu*, Dong Li, Tian Song

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Citation (Scopus)

Abstract

In view of the high complexity of pruning algorithm for OBS (optimal brain surgery) process and the deficiency of its match usage with training algorithm, this paper presents a penalty OBS computational model, in which the pruning condition is considered as a penalty term integrated in the objective function of NN (neural network). Based on its theoretical convergence, this model is realized by adopting the conjugate gradient method. Moreover, the effectiveness of this model is validated by a simulation test. The parallelization of network training process and OBS process ensures the accuracy and the efficiency of regularization so as to improve the generalization capacity of NN.

Original languageEnglish
Title of host publicationInformation and Automation - International Symposium, ISIA 2010, Revised Selected Papers
Pages48-57
Number of pages10
DOIs
Publication statusPublished - 2011
Event2010 International Symposium on Information and Automation, ISIA 2010 - Guangzhou, China
Duration: 10 Nov 201011 Nov 2010

Publication series

NameCommunications in Computer and Information Science
Volume86 CCIS
ISSN (Print)1865-0929

Conference

Conference2010 International Symposium on Information and Automation, ISIA 2010
Country/TerritoryChina
CityGuangzhou
Period10/11/1011/11/10

Keywords

  • conjugate gradient
  • convergence
  • model
  • network pruning
  • neural network
  • optimal brain surgery process

Fingerprint

Dive into the research topics of 'Penalty-optimal brain surgeon process and its optimize algorithm based on conjugate gradient'. Together they form a unique fingerprint.

Cite this