Neural Network Design without Learning
The paper deals with homogeneous neural networks of the feed-forward type. Contrary to the classical learning methods, the authors present a theoretical approach to a new method of neural network design. As an example, the designing of the neural network for ridges detection in interferometry images has been presented. To ensure the correctness of theoretical considerations, several results of experiments obtained with real images have been presented as well.
KeywordsNeural Network Soft Computing Neural Network Design Object Recognition System Hierarchical Neural Network
Unable to display preview. Download preview PDF.
- Gomolka Z. (2000),“Neural Networks for Interferometry Image Processing”, PhD thesis, AGH Cracow, (In Polish).Google Scholar
- Dudek-Dyduch E. (1986), „Cybernetics of Neuron like Systems, — Parallel Signal Processing; “, Polish Cybernetics Society Presss, Cracow (In Polish).Google Scholar
- Dudek-Dyduch E., Gomolka Z. (1996)“Adjusting parameters of two dimensional neural network”, Proc. Of the Second Conference of Polish Neural SocietyGoogle Scholar
- Dudek-Dyduch E., Dyduch T. (1996), “On feed forward homogenous layer neural nets”, Proc. Of the Second Conference of Polish Neural Society, Szczyrk.Google Scholar
- Dudek-Dyduch E., Gomolka Z. (2000), “Hybrid System–Wavelet Decomposition and Hierarchical Neural networks, Fifth International Conference Neural Networks and Soft Computing- Zakopane, June 6–10 2000, 186–195.Google Scholar
- Gawronski R. (1975), „Bionic Problems in the Wide Systems“”, Department of National Defense Presss, Warsaw1975, (In Polish).Google Scholar
- Klette R., Zamperoni P. (1996), “Handbook of image processing operators”, John Wileysons 1996.Google Scholar
- Dyduch T., Gomolka Z. (2000), “Implementation of homogeneous neural network on highly parallel CNAPS platform”, Fifth International Conference Neural Networks and Soft Computing- Zakopane, June 6–10 2000, 186–195.Google Scholar