Generalized Variable-Kernel Similarity Metric Learning

  • Johannes J. Naudé
  • Michaël A. van Wyk
  • Barend J. van Wyk
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 3138)


Proximity-based classifiers such as RBF-networks andnearest-neighbour classifiers are notoriously sensitive to the metric used to determine distance between samples. In this paper a method for learning such a metric from training data is presented. This algorithm is a generalization of the so called Variable-Kernel Similarity Metric (VSM) Learning, originally proposed by Lowe and is therefore known as Generalized Variable-Kernel Similarity Metric (GVSM) learning. Experimental results show GVSM to be superior to VSM for extremely noisy or cross-correlated data.


Generalization Performance Neighbour Method Cross Validation Error Query Vector Training Observation 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


  1. 1.
    Lowe, D.G.: Similarity Metric Learning for a Variable-Kernel Classifier. Neural Computation 7(1), 72–85 (1995)CrossRefGoogle Scholar
  2. 2.
    Cover, T.M., Hart, P.E.: Nearest neighbour pattern classifification. IEEE Transactions on Information Theory 13(1), 21–27 (1967)zbMATHCrossRefGoogle Scholar
  3. 3.
    Duda, R.O., Hart, P.E.: Pattern Classification and Scene Analysis, New York. Wiley, Chichester (1973)zbMATHGoogle Scholar
  4. 4.
    Friedman, J.H., Bentley, J.L., Finkel, R.A.: An algorithm for finding best matches in logarithmic expected time. ACM Trans. Math. Software 3, 209–226 (1973)CrossRefGoogle Scholar
  5. 5.
    Sproull, R.F.: Refinements to nearest-neighbour searching in k−d trees. Algorithmica 6, 579–589 (1973)CrossRefMathSciNetGoogle Scholar
  6. 6.
    Beis, J.S., Lowe, D.G.: Shape indexing using approximate nearest neighbour in high-dimensional spaces. In: Conference on Computer Vision and Pattern Recognition, Puerto Rico, pp. 1000–1006 (1997)Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2004

Authors and Affiliations

  • Johannes J. Naudé
    • 1
    • 2
  • Michaël A. van Wyk
    • 3
  • Barend J. van Wyk
    • 3
  1. 1.Rand Afrikaans UniversityJohannesburgSouth Africa
  2. 2.Kentron DynamicsCenturionSouth Africa
  3. 3.French South-African Technical Institute in ElectronicsTshwane University of TechnologyPretoriaSouth Africa

Personalised recommendations