95 articles 

inria-00104833, version 1

A Fast Learning Strategy Using Pattern Selection for Feedforward Neural Networks

Szilárd Vajda () 1, Yves Rangoni () 1, Hubert Cecotti () 1, Abdel Belaïd () 1

Tenth International Workshop on Frontiers in Handwriting Recognition 2006 - IWFHR'10 (2006) 6

Abstract: Intelligent pattern selection is an active learning strategy where the classifiers select during training the most informative patterns. This paper investigates such a strategy where the informativeness of a pattern is measured as the approximation error produced by the classifier. The algorithm builds the training corpus starting from a small randomly chosen initial dataset and new patterns are added to the learning corpus based on their error sensitivity. The training dataset expansion is based on the selection of the most erroneous patterns. Our experimental results on MNIST 1 separated digit dataset show that only 3.26%of training data are sufficient for training purpose without decreasing the performance (98.36%) of the resulting neural classifier.

  • 1:  READ (LORIA)
  • INRIA – CNRS : UMR7503 – Université Henri Poincaré - Nancy I – Université Nancy II – Institut National Polytechnique de Lorraine (INPL)
  • Domain : Computer Science/Computer Vision and Pattern Recognition
    Computer Science/Document and Text Processing
  • Keywords : Pattern selection – incremental learning – handwritten digit recognition – neural networks – support vector machines
  • Comment : http://www.suvisoft.com
 
  • inria-00104833, version 1
  • oai:hal.inria.fr:inria-00104833
  • From: 
  • Submitted on: Monday, 9 October 2006 14:59:40
  • Updated on: Wednesday, 8 December 2010 17:24:19