Neuronaler Netzwerk-Algorithmus für LDA/GSVD

The performance of classical linear discriminant analysis based on generalized singular value decomposition (LDA/GSVD) deteriorates when dealing with unlabeled datasets, as LDA requires predefined inputs and targets. Furthermore, the LDA/GSVD algorithm suffers from high computational costs due to its complex mathematical calculations and iterations. To address these issues, this study introduces the self-organizing map (SOM) as a new method for labeling datasets and the development of an algorithm based on artificial neural networks to overcome the computational costs of LDA/GSVD. The results show that the use of SOM and ANN effectively resolves the problems of the traditional LDA/GSVD algorithm.