Sains Ma1aysiana 26(3&4): 15-24 (1997)                                                                                    Pengajian Kuantitatif/

                                                                                                                                                                Quantitative Studies

Parallel Simulation for Solving Intertwined Spiral

Problem Using NEUCOMP2

Md. Nasir Bin Sulaiman

Department of Computer Science

Universiti Putra Malaysia

43400 UPM Serdang Selangor D.E.

D.J. Evans

Parallel Algorithms Research Centre

Loughborough University

Leicestershire LEI I 3TU

United Kingdom.

ABSTRACT

NEUCOMP2 is a parallel Neural Network Compiler for a shared-memory parallel machine. It compiles a program written as a list of mathematical specifications of Neural Network (NN) models and then translates it into a chosen target program which contains parallel codes. In this paper we present the performance results for intertwined spiral problem on popular NN models. The models are the Backpropagation, Kohonen, and Counter propagation models. NEUCOMP2 was developed on the Sequent 8000 computer system at PARC.

ABSTRAK

NEUCOMP2 adalah Pengkompilasi Rangkaian Neuron selari untuk mesin selari berkongsi-ingatan. Ia mengkompilasi atur cara yang ditulis dalam bentuk senarai spesifikasi matematik bagi model Rangkaian Neuron (RN) dan kemudian menterjemahkannya kepada atur cara sasaran yang dipilih mengandungi kod-kod selari. Dalam makalah ini dibentangkan keputusan prestasi penyelesaian masalah lingkaran berkembar menggunakan model­-model RN tertentu. Model-model ini adalah model rangkaian Perambatan Balik, Kohonen, dan Lawan Perambatan. EUCOMP2 dibangunkan pada sistem komputer Sequent 8000 di PARC.

RUJUKAN/REFERENCES

Beale, R. & Jackson, T. 1990. Neural computing: An introduction. A. Hilger.
Dayhoff, J. E. 1990. Neural network architecture: An introduction. New York: Van Nostrand Reinhold.

Evans, D. J & Sanossian, H. Y. Y. 1993. A gradient range heuristic algorithm for backpropagation. J. of Microcomputer Applications 16: 179-188.

Evans, D. J. & Sulaiman, M. N. 1994. NEUCOMP- A NEURAL NETWORK COMPILER. International Journal of Computer Math. 53: 29-44.

Evans, D. J. & Sulaiman, M. N. 1995. A Neural Network Computer Simulation to the           Intertwined Spiral Problem. Workshop - Research Network on Nantechnological and Holographics Methods for real-time Pattern Recognition (NATHAN), Berlin.

Lang, KJ. & Witbrock, M. J. 1988. Learning to tell two spirals apart. Proceedings of the Connectionist models, Summer School, Morgan-Kaufman, 52-59

Leighton, R. & Wieland, A. 1992. The Aspirin/MIGRANES software tools user's manual. Washington: MITRE Corporation.

Osterhaug, A. 1989. Guide to Parallel programming. Sequent Computer Systems, Inc.        

Shumsheruddin, D. 1992. The neural network paradigm. In Kronsjo, L. & Shumsheruddin, D. (ed.). Advanced topics in computer series. Advances in parallel algorithms, 66-84. Blackwell Scientific Publication.     

Sulaiman, M.N. & Evans, D. J. 1994. NEUCOMP2 - PARALLEL NEURAL NETWORK COMPILER. Technical report Department of Computer Studies. Loughborough University of Technology. Computer Studies 933.