[Neural network algorithms for recognition of complex binary images without their previous segmentation].
The authors propose a new neural network algorithm for binary images. It is based on a binary version of the Grossberg ART-architecture. Its changes made in the mechanism responsible for the competition of categorical neurones are rolled on the neurophysiological findings indicative of the fact that there are intercalary neurons in the inhibitory feedback chains. The proposed scheme in making the competition enables one to supplement the ability to decompose a complex input pattern into the basis (which is frequently non-orthogonal) of previously stored images to the advantages of ART-architectures.
Algorithms↗