On Not Making Dissimilarities Euclidean

  • Elzbieta Pękalska
  • Robert P. W. Duin
  • Simon Günter
  • Horst Bunke
Part of the Lecture Notes in Computer Science book series (LNCS, volume 3138)


Non-metric dissimilarity measures may arise in practice e.g. when objects represented by sensory measurements or by structural descriptions are compared. It is an open issue whether such non-metric measures should be corrected in some way to be metric or even Euclidean. The reason for such corrections is the fact that pairwise metric distances are interpreted in metric spaces, while Euclidean distances can be embedded into Euclidean spaces. Hence, traditional learning methods can be used.

The k-nearest neighbor rule is usually applied to dissimilarities. In our earlier study [2,3], we proposed some alternative approaches to general dissimilarity representations (DRs). They rely either on an embedding to a pseudo-Euclidean space and building classifiers there or on constructing classifiers on the representation directly. In this paper, we investigate ways of correcting DRs to make them more Euclidean (metric) either by adding a proper constant or by some concave transformations. Classification experiments conducted on five dissimilarity data sets indicate that non-metric dissimilarity measures can be more beneficial than their corrected Euclidean or metric counterparts. The discriminating power of the measure itself is more important than its Euclidean (or metric) properties.


Euclidean Space Average Error Negative Eigenvalue Handwritten Digit Deformable Template 
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.
    Blake, C.L., Merz, C.J.: UCI repository of machine learning databases. Dept. of Information and Computer Sciences. University of California, Irvine (1998)Google Scholar
  2. 2.
    Bunke, H., Günter, S., Jiang, X.: Towards bridging the gap between statistical and structural pattern recognition: Two new concepts in graph matching. In: Singh, S., Murshed, N., Kropatsch, W.G. (eds.) ICAPR 2001. LNCS, vol. 2013, pp. 1–11. Springer, Heidelberg (2001)CrossRefGoogle Scholar
  3. 3.
    Chow, C.K.: On optimum recognition error and reject tradeoff. IEEE Trans. on Information Theory IT-16(1), 41–46 (1970)zbMATHCrossRefGoogle Scholar
  4. 4.
    Courrieu, P.: Straight monotonic embedding of data sets in Euclidean spaces. Neural Networks 15, 1185–1196 (2002)CrossRefGoogle Scholar
  5. 5.
    Goldfarb, L.: A new approach to pattern recognition. In: Progress in Pattern Recognition, vol. 2, pp. 241–402. Elsevier Science Publishers B.V, Amsterdam (1985)Google Scholar
  6. 6.
    Gower, J.C.: Metric and Euclidean Properties of Dissimilarity Coefficients. Journal of Classification 3, 5–48 (1986)zbMATHCrossRefMathSciNetGoogle Scholar
  7. 7.
    Jacobs, D.W., Weinshall, D., Gdalyahu, Y.: Classification with Non-Metric Distances: Image Retrieval and Class Representation. IEEE Trans. on PAMI 22(6), 583–600 (2000)Google Scholar
  8. 8.
    Jain, A.K., Zongker, D.: Representation and recognition of handwritten digits using deformable templates. IEEE Trans. on PAMI 19(12), 1386–1391 (1997)Google Scholar
  9. 9.
    Moreno-Seco, F., Micó, L., Oncina, J.: A modification of the LAESA algorithm for approximated k-nn clasification. Pattern Recogn. Letters 24(1-3), 47–53 (2003)zbMATHCrossRefGoogle Scholar
  10. 10.
    Dubuisson, M.P., Jain, A.K.: Modified Hausdorff distance for object matching. In: 12th Int. Conf. on Pattern Recognition, vol. 1, pp. 566–568 (1994)Google Scholar
  11. 11.
    Pękalska, E.: working title: Dissimilarity-based pattern recognition. PhD thesis, Delft University of Technology, The Netherlands, to appear (2004)Google Scholar
  12. 12.
    Pękalska, E., Duin, R.P.W.: Dissimilarity representations allow for building good classifiers. Pattern Recogn. Letters 23(8), 943–956 (2002)zbMATHCrossRefGoogle Scholar
  13. 13.
    Pękalska, E., Paclík, P., Duin, R.P.W.: A Generalized Kernel Approach to Dissimilarity Based Classification. J. of Mach. Learn. Research 2, 175–211 (2001)CrossRefGoogle Scholar
  14. 14.
    Roth, V., Laub, J., Buhmann, J.M., Müller, K.-R.: Going metric: Denoising pairwise data. In: Advances in NIPS 15, pp. 841–856. MIT Press, Cambridge (2003)Google Scholar
  15. 15.
    Wilson, C.L., Garris, M.D.: Handprinted character database 3. Technical report, National Institute of Standards and Technology (February 1992)Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2004

Authors and Affiliations

  • Elzbieta Pękalska
    • 1
  • Robert P. W. Duin
    • 1
  • Simon Günter
    • 2
  • Horst Bunke
    • 2
  1. 1.ICT Group, Faculty of Electrical Engineering, Mathematics and Computer SciencesDelft University of TechnologyThe Netherlands
  2. 2.Department of Computer ScienceUniversity of BernSwitzerland

Personalised recommendations