A parallelization method for neural networks with weak connection design

  • Alexandra I. Cristea
  • Toshio Okamoto
VII Poster Session Papers
Part of the Lecture Notes in Computer Science book series (LNCS, volume 1336)


Hereby we present the construction and usage of “Weak Connectiou”(WeCo) on Neural Networks(NN). We will show how these parallelization hypothesis increases the final system flexibility. The net design is based on standard procedures, but changed accordingly to WeCo parallelization principles. WeCo means parallelization with less weight on communication systems, as in: fine, medium and coarse grain parallelism, or between the parts of the implementation program. WeCo lays in-between parallel computers and sequential machines, building the bridge between them.


Neural Networks Weak Connections Parallelism Access Points 


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    D.Dasgupta and D.R.McGregor,1992. Designing Application-Specific Neural Networks using the Structured Genetic Algorithm. Proceedings of COGANN-92 (Internat. Workshop on Comb. of Genetic Alg. and NN, June6,1992, Ed.Whitley and Schaffer, Publ.: IEEE Computer Society Press).Google Scholar
  2. 2.
    D.Komo, C.I.Chang and H.Ko,1994. Neural Network Technology for Stock Market Index Prediction. International Symposium on Speech, Image Processing and Neural Networks, 13–16 April 1994, HongKong, IEEE, pages 543–546.Google Scholar
  3. 3.
    M.Lam. Software Pipelining: An Effective Scheduling Technique for VLIW Machines. In Proc. ACM SIGPLAN, pp.318–328, l988.Google Scholar
  4. 4.
    Daniel E. Lenoski The Design and Analysis of Dash: A Scalable Directory-based Multiprocessor. PhD Thesis, Stanford University, CSL, 1992.Google Scholar
  5. 5.
    A.U.Levin, T.K.Leeu,J.E.Moody,1994. Fast Pruning Using Principal Components. Advances in Neural Information Processing 6, J.Cowau, G.Tesauro, J.Alspector, eds., Morgan Kaufmann, San Mateo, CA, 1994.Google Scholar
  6. 6.
    S. Mori and et al. A Distributed Shared Memory Multiprocessor: ASURA — Memory and Cache Architectures-. In Proc. Supercomputing 99, pp.740–749, 1993Google Scholar
  7. 7.
    N. Murata, S. Yoshizawa and S. Amari,1995. Network Information Criterion-Determining the Number of Hidden Units for an Artificial NN Model. Dept. of Mathematical Eng. and Info. Physics, Fac. of Engineering, Univ. of Tokyo,ftp-source.Google Scholar
  8. 8.
    M.E.Nelson, W.Furmanski, J.M.Bower,1989. Simulating Neurons and Networks on Parallel Computers. Methods in Neuronal Modeling, From Synapses to Networks, ed. C.Koch and I.Segev, MIT, 1989,chapt. 12.Google Scholar
  9. 9.
    C. Brownhill, A. Nicolau, S. Novack and C. Polychronopoulos The Promis Compiler. In Proc. PACT97, will appear, 1997Google Scholar
  10. 10.
    G.Thimm, E.Fiesler,1996. A Neural Network Construction Method based an Boolean Logic. accepted for IEEE International Conference on Tools with Artificial Intelligence proceedings, Toulose, France, 1996.Google Scholar
  11. 11.
    G.Thimm, E.Fiesler,1996. Neural Network Pruning and Pruning Parameters, 1st OWSC. http://www/, 1st Online Workshop on Soft Computing, Nagoya. Japan, 1996.Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 1997

Authors and Affiliations

  • Alexandra I. Cristea
    • 1
  • Toshio Okamoto
    • 1
  1. 1.Graduated School of Information SystemsUniversity of Electro-CommunicationTokyoJapan

Personalised recommendations