Abstract
A massive advancement in image processing technologies supports a rapid on-line classification in well-defined pattern recognition tasks. Yet in general, there remains an urgent need for robust pre-processing techniques applicable independent of the task to be solved.
Currently, two such approaches based on artificial neural networks are known from the literature, namely the so-called Extreme Learning Machines and the convolutional neural networks. In particular, the convolutional networks contain feature-detecting layers trained along with the rest of the network by means of the back-propagation algorithm.
Another alternative for an automatic design and training of such pre-processors represents self-organization. Its main advantage consists in the much faster convergence during training.
This work describes and compares several models of self-organizing networks with respect to their use for feature extraction and unsupervised building of preprocessing stage.