Representation and Evolution of Neural Networks

  • Martin Mandischer


An evolutionary approach for developing improved neural network architectures is presented. It is shown that it is possible to use genetic algorithms for the construction of backpropagation networks for real world tasks. Therefore a network representation is developed with certain properties. Results with various application are presented.


Genetic Algorithm Network Representation Good Network Standard Network Real World Task 
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.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. [1]
    D. Fox, V. Heinze, K. Möller, S. Thrun, and G. Veenker. Learning by error-driven decomposition. In T. Kohonen, K. Mäkisara, O. Simula, and J. Kangas, editors, Artificial Neural Networks, volume 1, pages 207–212, Amsterdam, June 1991. North-Holland.Google Scholar
  2. [2]
    D. E. Goldberg. Genetic Algorithms in Search, Optimization and Machine Learning. Addison-Wesley, 1989.Google Scholar
  3. [3]
    S. A. Harp, T. Samad, and A. Guha. Designing application-specific neural networks using the genetic algorithm. In D. S. Touretzky, editor, Proceedings of IEEE conference on Neural Information Processing Systems, volume 2, pages 447–454, San Mateo, 1990. (Denver 1989), Morgan Kaufmann.Google Scholar
  4. [4]
    J. S. Judd. Neural Network Design and the Complexity of Learning. MIT Press — Bradford Book, Cambridge, MA, 1990.Google Scholar
  5. [5]
    H. J. Lin and J. S. Vitter. Complexity issues in learning by neural networks. Technical Report TR-CS-90-01, Department of Computer Science, Brown University, Providence, II, 1990.Google Scholar
  6. [6]
    M. Mandischer. Genetische Algorithmen zur Optimierung Konnektionistischer Modelle. Master’s thesis, Department of Computer Science, University of Dortmund, Dortmund, PO BOX 500500, February 1992.Google Scholar
  7. [7]
    G. F. Miller, P. M. Todd, and S. U. Hegde. Designing neural networks using genetic algorithms. In J. D. Schaffer, editor, Proceedings of the Third International Conference on Genetic Algorithms, pages 379-384, San Mateo, 1989. (Arlington 1989), Morgan Kaufmann.Google Scholar
  8. [8]
    T. Rohne. Künstliche neuronale netze zur eckendetektion in digitalisierten grauwertbildern. Master’s thesis, University of Dortmund, Department of Computer Science X, Dortmund, FRG, 1991.Google Scholar
  9. [9]
    N. Schaudolph and R. K. Belew. Dynamic parameter encoding for genetic algorithms. Technical Report CSE-TR#CS90-175, University of California, San Diego, CA, 1990.Google Scholar
  10. [10]
    W. Schiffmann, J. Merten, and W. Randolf. Performance evaluation of evolutionary created neural network topologies. In First International Workshop on Parallel Problem Solving from Nature, pages A-III:-1-11, Dortmund, FRG, 1990. University of Dortmund, Department of Computer Science X.Google Scholar
  11. [11]
    A. Ultsch, R. Hannuschka, U. Hartmann, M. Mandischer, and V. Weber. Optimizing symbolic proofs with connectionist models. In T. Kohonen, K. Mäkisara, O. Simula, and J. Kangas, editors, Artificial Neural Networks, volume 1, pages 585–590, Amsterdam, June 1991. North-Holland.Google Scholar
  12. [12]
    G. Weiß. Combining neural and evolutionary learning: Aspects and approaches. Technical Report FKI-132-90, Technische Universität München, 1990.Google Scholar
  13. [13]
    H. White and K. Hornik. Universal approximation using feedforward networks with nonsigmoid hidden layer activation functions. In Proceedings of the International Joint Conference on Neural Networks. (San Diego, CA), IEEE, 1989.Google Scholar

Copyright information

© Springer-Verlag/Wien 1993

Authors and Affiliations

  • Martin Mandischer
    • 1
  1. 1.Department of Computer Science VIUniversity of DortmundGermany

Personalised recommendations