A modified spreading algorithm for autoassociation in weightless neural networks
This paper describes a problem with the conventional Hamming metric spreading algorithm often employed in weightless neural networks. The algorithm can cause incorrect classifications in some cases where a section of the neuron input vector is noise. The conditions under which such error occurs are described and a modified spreading algorithm proposed to overcome this problem has its validity demonstrated theoretically and tested in two practical applications.
Unable to display preview. Download preview PDF.
- C. Browne and I. Aleksander, Digital general neural units with controlled transition probabilities Electronics Letters, Vol. 32, No. 9, pp 824–825, 1996.Google Scholar
- I. Aleksander, Neural Systems Engineering: towards a unified design discipline? IEE J. Computation and Control, Vol.6, pp. 259–265, 1990.Google Scholar
- I. Aleksander, T. J. W. Clarke, and Braga A. P. Binary neural systems: A unified approach to their analysis and design.Google Scholar
- J. D. L. P. Castro. Acquiring Semantic Power in “Weightless Neural Network. In 1995 IEEE International Conference on Neural Networks Proceedings, pages 674–679.Google Scholar
- N. J. Sales and R. G. Evans. An approach to solving the symbol grounding problem: Neural networks for object naming and retrieval. In Proc. CMC-95, 1995.Google Scholar
- P. Ntourntoufis. Aspects of the theory of weightless artificial neural networks. PhD thesis, Imperial College of Science, Technology and Medicine, 1994.Google Scholar
- R Evans. Reconstructing quatisation-distorted images with a Neural State Machine. 1994.Google Scholar