Network of adaptive-resonance theory with multilevel memory

Author(s):  D.G. Bukhanov, Belgorod State Technological University named after V.G. Shukhov, Belgorod, Russia

V.M. Polyakov, candidate of Sciences, associate Professor, Belgorod State Technological University named after V.G. Shukhov, Belgorod, Russia

Issue:  Volume 45, № 4

Rubric:  Computer simulation history

Annotation:  The analysis of the artificial neural network based on the adaptive-resonance theory with continuous input signals is described in the paper. Reviewed general shortcomings of the network and the problems with its using. The main problem is low search speed of the active resonating memory neuron in the field F2, which leads to less accurate recognition of objects in real-time systems. To solve the problem, a new model of the recognition field F2 of the ART-2 network is proposed, which is a tree structure, with a recurrently changing similarity parameter for each subsequent level. At each level, the similarity parameter increases, that leads to a sequential search for an active resonating neuron. A comparative analysis of the time characteristics of the proposed network and the classical implementation is carried out. The advantages of the proposed network model based on the adaptive resonance theory are experimentally proved. Network with the proposed modification shows better speed than the classic ART-2 without loss of accuracy

Keywords:  artificial neural networks, adaptive resonance theory, clustering, data mining

Full text (PDF):  Download

Downloads count:  528