Abstract: New non-conventional system of the computational intelligence is proposed. It has growing structure similar to the Cascade-Correlation? Learning Architecture designed by S. E. Fahlman and C. Lebiere but differs from it in type of artificial neurons. Quadratic neurons are used as nodes in introduced architecture. These simple elements can be quickly adjusted using high-speed learning procedures. Proposed approach allows to reduce time required for weight coefficients adjustment and to decrease training dataset size in comparison with conventional neural networks. Also on-board realization of quadratic neuron is quite simple and therefore implementation of entire cascade architecture in hardware is very easy.

Keywords: artificial neural networks, constructive approach, quadratic neuron, real-time processing, online learning.

ACM Classification Keywords: I.2.6 Learning – Connectionism and neural nets.

Link:

THE CASCADE GROWING NEURAL NETWORK USING QUADRATIC NEURONS AND ITS LEARNING ALGORITHMS FOR ON-LINE INFORMATION PROCESSING

Yevgeniy Bodyanskiy, Yevgen Viktorov, Iryna Pliss

http://www.foibg.com/ibs_isc/ibs-13/ibs-13-p03.pdf