ReART: a Novel Adaptive Resonance Theory for Pattern Learning and Recognition based on NO Retrograde

A novel adaptive resonance theory (ART) based on nitric oxide (NO) retrograde mechanism, named retrograde ART (ReART), is presented in this paper. Within the theoretical framework of the searching process in ART 3, we analyze the transmitter release among chemical synapses, and propose a novel search hypothesis which incorporates angle and amplitude information to decide whether an external input matches the long-term memory (LTM) weights of an active node or not. By introducing NO retrograde mechanism, the dynamic search and mismatch-reset cycle of ART 3 is improved. To avoid the potential phenomenon of pattern excursion in the node growing process, the forgetting mechanism is constructed. By incorporating the matched nodes and abandoning the erroneous nodes, the novel algorithm optimizes the node growing mechanism. The following simulations indicate that the proposed model has perfect classification, faster convergence and excellent disturbance rejection capability

[1]  Solomon H. Snyder,et al.  Nitric oxide, a novel neuronal messenger , 1992, Neuron.

[2]  Graham Leedham,et al.  Reinforcement learning combined with a fuzzy adaptive learning control network (FALCON-R) for pattern classification , 2005, Pattern Recognit..

[3]  Stephen Grossberg,et al.  ART 3: Hierarchical search using chemical transmitters in self-organizing pattern recognition architectures , 1990, Neural Networks.

[4]  Stephen Grossberg,et al.  ART 2-A: An adaptive resonance algorithm for rapid category learning and recognition , 1991, Neural Networks.

[5]  Allon Guez,et al.  ART based adaptive pole placement for neurocontrollers , 1991, Neural Networks.

[6]  S. Grossberg,et al.  ART 2: self-organization of stable category recognition codes for analog input patterns. , 1987, Applied optics.

[7]  Stephen Grossberg,et al.  ARTMAP: supervised real-time learning and classification of nonstationary data by a self-organizing neural network , 1991, [1991 Proceedings] IEEE Conference on Neural Networks for Ocean Engineering.

[8]  Stephen Grossberg,et al.  Fuzzy ARTMAP: A neural network architecture for incremental supervised learning of analog multidimensional maps , 1992, IEEE Trans. Neural Networks.

[9]  I. Jolliffe.  Principal Component Analysis , 2002 .

[10]  R. Nicoll,et al.  Long-term potentiation--a decade of progress? , 1999, Science.

[11]  R. A. Baxter,et al.  Error propagation and supervised learning in adaptive resonance networks , 1991, IJCNN-91-Seattle International Joint Conference on Neural Networks.

[12]  Paul T. Kelly,et al.  Nitric Oxide Acts as a Postsynaptic Signaling Molecule in Calcium/Calmodulin-Induced Synaptic Potentiation in Hippocampal CA1 Pyramidal Neurons , 1999, The Journal of Neuroscience.

[13]  R. Fisher.  THE USE OF MULTIPLE MEASUREMENTS IN TAXONOMIC PROBLEMS , 1936 .

[14]  Stephen Grossberg,et al.  Adaptive pattern classification and universal recoding: II. Feedback, expectation, olfaction, illusions , 1976, Biological Cybernetics.

[15]  Stephen Grossberg,et al.  Fuzzy ART: Fast stable learning and categorization of analog patterns by an adaptive resonance system , 1991, Neural Networks.

[16]  S. Grossberg.  Adaptive Resonance Theory , 2006 .