Advertisement

Objective Functions for Topography: A Comparison of Optimal Maps

  • Geoffrey J. Goodhill
  • Terrence J. Sejnowski
Conference paper
Part of the Perspectives in Neural Computing book series (PERSPECT.NEURAL)

Abstract

Topographic mappings are important in several contexts, including data visualization, connectionist representation, and cortical structure. Many different ways of quantifying the degree of topography of a mapping have been proposed. In order to investigate the consequences of the varying assumptions that these different approaches embody, we have optimized the mapping with respect to a number of different measures for a very simple problem — the mapping from a square to a line. The principal results are that (1) different objective functions can produce very different maps, (2) only a small number of these functions produce mappings which match common intuitions as to what a topographic mapping “should” actually look like for this problem, (3) the objective functions can be put into certain broad categories based on the overall form of the maps, and (4) certain categories of objective functions may be more appropriate for particular types of problem than other categories.

Keywords

Objective Function Simulated Annealing Topographic Mapping Input Space Output Space 
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.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. [1]
    Barlow, H.B. (1989). Unsupervised learning. Neural Computation, 1, 295–311.CrossRefGoogle Scholar
  2. [2]
    Bezdek, J.C. and Pal, N.R. (1995). An index of topological preservation for feature extraction. Pattern Recognition, 28, 381–391.CrossRefGoogle Scholar
  3. [3]
    Cowey, A. (1979). Cortical maps and visual perception. Qua. Jou. Exper. Psychol., 31, 1–17.CrossRefGoogle Scholar
  4. [4]
    Durbin, R. and Mitchison, G. (1990). A dimension reduction framework for understanding cortical maps. Nature, 343, 644–647.CrossRefGoogle Scholar
  5. [5]
    Goodhill, G. J., Finch, S. and Sejnowski, T. J. (1995). Quantifying neighbourhood preservation in topographic mappings. Institute for Neural Computation Technical Report Series, No. INC-9505, November 1995. Available from http://www.giccs.georgetown.edu/—geoffGoogle Scholar
  6. [6]
    Goodhill, G.J. and Sejnowski, T.J. (1996) Quantifying neighbourhood preservation in topographic mappings. In: “Proceedings of the 3rd Joint Symposium on Neural Computation”, University of California, San Diego and California Institute of Technology, Vol. 6, Pasadena, CA: California Institute of Technology, 61–82. Available from http://www.giccs.georgetown.edu/ —geoffGoogle Scholar
  7. [7]
    Goodhill, G.J. and Sejnowski, T.J. (1997). A unifying objective function for topographic mappings. Neural Computation, 9, 1291–1304.Google Scholar
  8. [8]
    Hubel, D.H. and Wiesel, T.N. (1977). Functional architecture of the macaque monkey visual cortex. Proc. R. Soc. Lond. B 198 1-59.Google Scholar
  9. [9]
    Kirkpatrick, S., Gelatt, C.D. and Vecchi, M.P. (1983). Optimization by simulated annealing. Science, 220, 671–680.MATHCrossRefMathSciNetGoogle Scholar
  10. [10]
    Kohonen, T. (1982). Self-organized formation of topologically correct feature maps. Biol. Cybern., 43, 59–69.MATHCrossRefMathSciNetGoogle Scholar
  11. [11]
    Krzanowski, W.J. (1988). Principles of multivariate analysis: a user’s perspective. Oxford statistical science series; v. 3. Oxford University Press.Google Scholar
  12. [12]
    van Laarhoven, P.J.M. and Aarts, E.H.L. (1987). Simulated annealing: theory and applications. Reidel, Dordrecht, Holland.Google Scholar
  13. [13]
    Luttrell, S.P. (1990). Derivation of a class of training algorithms. IEEE Trans. Neural Networks, 1, 229–232.CrossRefGoogle Scholar
  14. [14]
    Mitchison, G. (1991). Neuronal branching patterns and the economy of cortical wiring. Proc. Roy. Soc. B., 245, 151–158.CrossRefGoogle Scholar
  15. [15]
    Mitchison, G. (1995). A type of duality between self-organizing maps and minimal wiring. Neural Computation, 7, 25–35.CrossRefGoogle Scholar
  16. [16]
    Nelson, M.E. and Bower, J.M. (1990). Brain maps and parallel computers. Trends Neurosci, 13, 403–408.Google Scholar
  17. [17]
    Sammon, J.W. (1969). A nonlinear mapping for data structure analysis. IEEE Trans. Comput., 18, 401–409.CrossRefGoogle Scholar
  18. [18]
    Shepard, R.N. (1980). Multidimensional scaling, tree-fitting and clustering. Science, 210, 390–398.MATHCrossRefMathSciNetGoogle Scholar
  19. [19]
    Torgerson, W.S. (1952). Multidimensional Scaling, I: theory and method. Psychometrika, 17, 401–419.MATHCrossRefMathSciNetGoogle Scholar
  20. [20]
    Udin, S.B. and Fawcett, J.W. (1988). Formation of topographic maps. Ann. Rev. Neurosci., 11, 289–327.Google Scholar

Copyright information

© Springer-Verlag London Limited 1998

Authors and Affiliations

  • Geoffrey J. Goodhill
    • 1
  • Terrence J. Sejnowski
    • 2
  1. 1.Georgetown Institute for Cognitive and Computational SciencesGeorgetown University Medical CenterUSA
  2. 2.The Salk Institute for Biological StudiesLa JollaUSA

Personalised recommendations