An Algorithm for Fast Pattern Recognition with Random Spikes

  • Udo A. Ernst
  • David Rotermund
  • Klaus R. Pawelzik
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 3175)


The human brain classifies natural scenes and recognizes objects in complex visual patterns with a high precision in a minimum amount of processing time. Only few action potentials (spikes) per neuron and per processing stage are sufficient to achieve this astonishingly high performance, despite the random nature of the incoming spike trains. In this contribution, we present a novel algorithm which updates the internal representation of patterns in a generative model with each incoming spike. We first demonstrate that our algorithm is capable of learning a suitable representation of pattern ensembles from stochastically generated spike trains. This representation is then used for classifying test patterns, requiring less than one spike per input node to achieve a performance comparable to standard algorithms in pattern recognition.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Bethge, M., Rotermund, D., Pawelzik, K.: Optimal short-term population coding: When Fisher information fails. Neural Computation 14(10), 2317–2351 (2002); A second order phase transition in neural rate coding: Binary encoding is optimal for rapid signal transmission. Phys. Rev. Lett. 90, 88104 (2002)Google Scholar
  2. 2.
    Gerstner, W.: Population Dynamics of Spiking Neurons: Fast Transients, Asynchronous States, and Locking. Neural Computation 12, 43–89 (2000)CrossRefGoogle Scholar
  3. 3.
    Lanteri, H., Roche, M., Aime, C.: Penalized maximum likelihood image restoration with positivity constraints: multiplicative algorithms. Inverse Problems 18, 1397–1419 (2002)zbMATHCrossRefMathSciNetGoogle Scholar
  4. 4.
    Lee, D.D., Seung, S.H.: Learning the parts of objects by non-negative matrix factorization. Nature 401, 788–791 (1999)CrossRefGoogle Scholar
  5. 5.
    Pawelzik, K.R., Rotermund, D., Ernst, U.A.: Building representations spike by spike. In: Elsner, N., Zimmermann, H. (eds.) Proceedings of the 29th Göttingen Neurobiology Conference, p. 1041. Georg Thieme Verlag, Stuttgart (2003)Google Scholar
  6. 6.
    Pawelzik, K.R., Ernst, U.A., Trenner, D., Rotermund, D.: Building representations spike by spike. In: Proceedings of the Society of Neuroscience Conference 2002, Orlando, p. 557.12 (2002)Google Scholar
  7. 7.
    Schölkopf, B.: Support Vector Learning, R. Oldenbourg Verlag, München (1997),
  8. 8.
    Thorpe, S., Fize, D., Marlot, C.: Speed of processing in the human visual system. Nature 381, 520–522 (1996)CrossRefGoogle Scholar
  9. 9.
    Thorpe, S., Delorme, A., van Rullen, R.: Spike-based strategies for rapid processing. Neural Networks 14, 521–525 (2001)CrossRefGoogle Scholar
  10. 10.
    Wiener, M., Richmond, B.J.: Decoding spike trains instant by instant using order statistics and the mixture-of-Poissons model. J. Neurosci. 23, 2394–2406 (2003)Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2004

Authors and Affiliations

  • Udo A. Ernst
    • 1
  • David Rotermund
    • 1
  • Klaus R. Pawelzik
    • 1
  1. 1.Inst. for Theoretical NeurophysicsUniversity BremenBremenGermany

Personalised recommendations