Fast Orthogonal Neural Networks
The paper presents a novel approach to the construction and learning of linear neural networks based on fast orthogonal transforms. The orthogonality of basic operations associated with the algorithm of a given transform is used in order to substantially reduce the number of adapted weights of the network. Two new types of neurons corresponding to orthogonal basic operations are introduced and formulas for architecture-independent error backpropagation and weights adaptation are presented.
KeywordsDiscrete Cosine Transform Discrete Fourier Transform Digital Signal Processing Conjugate Gradient Method Basic Operation
Unable to display preview. Download preview PDF.
- 1.Osowski, S.: Neural networks for information processing (in Polish) OWPW, Warsaw (2000)Google Scholar
- 2.Rutkowski, L.: Methods and techniques of artificial intelligence (in Polish) Polish Scientific Publishers PWN (2005)Google Scholar
- 4.Yatsymirskii, M.N.: Fast algorithms for the discrete cosine transformation. Comput. Maths Math. Phys. 33, 267270 (1993)Google Scholar
- 6.Jacymirski, M., Szczepaniak, P.S.: Neural realization of fast linear filters. In: Proc. of the 4th EURASIP - IEEE Region 8 International Symposium on Video/Image Processing and Multimedia Communications, pp. 153–157 (2002)Google Scholar
- 7.Szczepaniak, P.S.: Intelligent computations, fast transforms and classifiers (in Polish) EXIT Academic Publishing House, Warsaw (2004)Google Scholar
- 8.Rabiner, L.R., Gold, B.: Theory and application of digital signal processing. Prentice-Hall, Englewood Cliffs (1975)Google Scholar
- 9.Jacymirski, M.: Fast homogeneous algorithms of cosine transforms, type II and III with tangent multipliers, pp. 727–741. AGH University of Science and Technology Press, Cracow (2003)Google Scholar
- 10.Yatsymirskyy, M.M.: Shifted in the time and frequency domains cosine and sine transforms fast algorithms with homogeneous structure (in Russian), Kiev. Izvestiya Vysshikh Uchebnykh Zavedenii, Radioelektronika, vol. 43, pp. 66–75 (2000)Google Scholar