Neural network experiences between perceptrons and support vectors

Robert P.W. Duin and Dick de Ridder
Pattern Recognition Group, Faculty of Applied Physics
Delft University of Technology, The Netherlands
duin@tn.tudelft.nl

Abstract:

After years of hesitation neural networks are now widely accepted as important tools for pattern recognition. As their characteristics differ strongly from the traditional techniques, neural networks have had a great influence on the recent developments of non-neural classifiers as well. A summary will be presented of the insights and fruits resulting from the neural network hype with an emphasis on the recently developed support vector machines.

• 1 The Challenge
• 2 Neural Network Properties
• 3 Further Observations
• 4 Experiments
• 5 Discussion