Experiments with a featureless approach to pattern recognition

Traditionally automatic pattern recognition is based on learning from examples of objects represented by features. In some applications it is hard to define a small, relevant set of features. At the cost of large learning sets and complicated learning systems discriminant functions have to be found. In this paper we discuss the possibility to construct classifiers entirely based on distances or similarities, without a relation with the feature space. This is illustrated by a number of . experiments based on the support object classifier Duin et al., 1997 , a derivative of Vapnik's support ˝ector classifier .Cortes and Vapnik, 1995 . q 1997 Elsevier Science B.V.

[1]  Robert P. W. Duin,et al.  Stabilizing classifiers for very small sample sizes , 1996, Proceedings of 13th International Conference on Pattern Recognition.

[2]  Anil K. Jain,et al.  A modified Hausdorff distance for object matching , 1994, Proceedings of 12th International Conference on Pattern Recognition.

[3]  Calyampudi R. Rao.  Handbook of statistics , 1980 .

[4]  R. Duin.  Small sample size generalization , 1995 .

[5]  Françoise Fogelman-Soulié,et al.  Multi-Modular Neural Network Architectures: Applications in Optical Character and Human Face Recognition , 1993, Int. J. Pattern Recognit. Artif. Intell..

[6]  Dick de Ridder,et al.  Shared Weights Neural Networks in Image Analysis , 1996 .

[7]  Anil K. Jain,et al.  39 Dimensionality and sample size considerations in pattern recognition practice , 1982, Classification, Pattern Recognition and Reduction of Dimensionality.

[8]  Robert P. W. Duin,et al.  Expected classification error of the Fisher linear classifier with pseudo-inverse covariance matrix , 1998, Pattern Recognit. Lett..