Abstract
A parallel mapping of self-organizing map (SOM) algorithm is presented for a partial tree shape neurocomputer (PARNEU). PARNEU is a general purpose parallel neurocomputer that is designed for soft computing applications. Practical scalability and a reconfigurable partial tree network are the main architectural features. The presented neuron parallel mapping of SOM with on-line learning illustrates a parallel winner neuron search and a coordinate transfer that are performed in the partial tree network. Phase times are measured to analyse speedup and scalability of the mapping. The performance of the learning phase in SOM with a four processor PARNEU configuration is about 26 MCUPS and the recall phase performs 30 MCPS. Compared to other mappings done for general purpose neurocomputers, PARNEU's performance is very good.
Similar content being viewed by others
References
Kohonen, T.: The Self-Organizing Maps, Springer-Verlag, Berlin, 1995.
Kohonen, T.: Self-Organization and Associative Memory, Springer-Verlag, Berlin, 1980.
Hammerström, D. and Nguyen, N.: An implementation of Kohonen feature maps on the adaptive solutions neurocomputer, in: T. Kohonen et. al. (eds), Artificial Neural Networks, Vol. 1, Proceedings of the ICANN-91, Espoo, Finland, 1991, pp. 715–720.
Cornu, T., Ienne, P., Neibur, D., Thiran, P. and Viredaz, M.: Design, implementation and test of a multi-model systolic neural network accelerator, Scient. Program. 5 (1996), 47–61.
Nordström, T.: Highly Parallel Computers for Artificial Neural Networks, Ph.D. Thesis, Luleå University of Technology, Luleå , Sweden, March 1995.
Nordström, T. and Svensson, B.: Using and designing massively parallel computers for artificial neural networks, J. Parallel Distrib. Comp. 14 (1992), 260–285.
Müller, U., Gunzinger, A. and Guggenbühl, W.: Fast neural net simulation with a DSP processor array, IEEE Trans. Neural Networks, 6 (1995), 203–213.
Morgan, N., Beck, J., Kohn, J., Bilmes, J., Allman, E. and Beer, J.: The ring array processor: A multiprocessing peripheral for connectionist applications, J. Parallel Distrib. Comput. 14 (1992), 248–259.
Ramacher, U.: SYNAPSE-A neurocomputer that synthesizes neural algorithms on a parallel systolic engine, J. Parallel Distrib. Comp. 14 (1992), 306–318.
Hämäläinen, T., Klapuri, H., Saarinen, J. and Kaski, K.: Mapping of SOM and LVQ algorithms on a tree shape parallel computer system, Parallel Comput. 23 (1997), 271–289.
Thiran, P., Peiris, V., Heim, P. and Hochet, B.: Quantization effects in digitally behaving circuit implementations of Kohonen networks, IEEE Trans. Neural Networks, 5 (1994), 450–458.
Rüping, S., Porrmann, M. and Rückert, U.: SOM accelerator system, Neurocomputing 21 (1998), 31–50.
Kolinummi, P., Hämälaäinen, T. and Kaski, K.: Designing a digital neurocomputer, IEEE Circ. Devices Mag. March 1997, 19–27.
Kolinummi, P., Hämaäläinen, P., Hämäläinen, T. and Saarinen, J.: PARNEU: general-purpose partial tree computer, Microprocessors and Microsystems 24 (2000), 23–42.
The Programmable Logic Data book, CA, USA, Xilinx Inc., 1994.
ADSP-2106x SHARC User's Manual, second edition, 7/96, Analog Devices Inc., USA.
Kolinummi, P., Hämäläinen T. and Kaski, K.: Mappings of SOMand LVQ on the partial tree shape neurocomputer, Inter. Conf. Neural Networks (ICNN'97), Vol. 2, Houston, TX, USA, June 1997, pp. 904–909.
Card, H. C., Rosendahl, G. K., McNeill, D. K. and McLeod, R. D.: Competitive learning algorithms and neurocomputer architecture, IEEE Trans. Comp. 47 (1998), 847–858.
Myklebust, G., Solheim, J. and Steen, E.: Speeding up small sized self organizing maps for use in visualization of multispectral medical images, Eighth IEEE Symp. on Computer-based Medical Systems, 1995, pp. 103–110.
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Kolinummi, P., Pulkkinen, P., Hämäläinen, T. et al. Parallel Implementation of Self-Organizing Map on the Partial Tree Shape Neurocomputer. Neural Processing Letters 12, 171–182 (2000). https://doi.org/10.1023/A:1009665814041
Issue Date:
DOI: https://doi.org/10.1023/A:1009665814041