GENERALIZATION PERFORMANCE ANALYSIS BETWEEN FUZZY ARTMAP AND GAUSSIAN ARTMAP NEURAL NETWORK

This paper examines the generalization characteristic of Gaussian ARTMAP (GAM) neural network in classification tasks. GAM performance for classification during training and testing is evaluated using the k-folds cross validation technique. A comparison is also done between GAM and Fuzzy ARTMAP (FAM) neural network. It is found that GAM performs better (98-99%) when compared to FAM (79-82%) using two different types of dataset. The difference between GAM and FAM is that input data to be to classified using FAM must be normalized in prior. Hence, three different normalization techniques are examined namely; unit range (UR), improved unit range (IUR) and improve linear scaling (ILS). This paper also proposes an alternative technique to search the best value for gamma γ parameter of GAM neural network, known as gamma threshold. A small number of training required for GAM also shows that its fundamental architecture retain the attractive parallel computing and fast learning properties of FAM.

[1]  B. W. Jervis,et al.  Integrated probabilistic simplified fuzzy ARTMAP , 2004 .

[2]  Carlos R. Minussi,et al.  Electric load forecasting using a fuzzy ART&ARTMAP neural network , 2005, Appl. Soft Comput..

[3]  James R. Williamson,et al.  Gaussian ARTMAP: A Neural Network for Fast Incremental Learning of Noisy Multidimensional Maps , 1996, Neural Networks.

[4]  Georgios C. Anagnostopoulos,et al.  Hypersphere ART and ARTMAP for unsupervised and supervised, incremental learning , 2000, Proceedings of the IEEE-INNS-ENNS International Joint Conference on Neural Networks. IJCNN 2000. Neural Computing: New Challenges and Perspectives for the New Millennium.

[5]  Roland Sauerbrey,et al.  Biography , 1992, Ann. Pure Appl. Log..

[6]  Michael Georgiopoulos,et al.  Boosted ARTMAP , 1998, 1998 IEEE International Joint Conference on Neural Networks Proceedings. IEEE World Congress on Computational Intelligence (Cat. No.98CH36227).

[7]  Georgios C. Anagnostopoulos,et al.  Overtraining in fuzzy ARTMAP: Myth or reality? , 2001, IJCNN'01. International Joint Conference on Neural Networks. Proceedings (Cat. No.01CH37222).

[8]  Michael Georgiopoulos,et al.  Hierarchical ARTMAP , 2000, Proceedings of the IEEE-INNS-ENNS International Joint Conference on Neural Networks. IJCNN 2000. Neural Computing: New Challenges and Perspectives for the New Millennium.

[9]  Yannis A. Dimitriadis,et al.  Artmap: Use of Mutual Information for Category Reduction in Fuzzy Artmap , 2002 .

[10]  Stephen Grossberg,et al.  Fuzzy ARTMAP: an adaptive resonance architecture for incremental learning of analog maps , 1992, [Proceedings 1992] IJCNN International Joint Conference on Neural Networks.

[11]  Ethem Alpaydin,et al.  Constructive Feedforward ART Clustering Networks — Part I , 2001 .

[12]  Ethem Alpaydin,et al.  Constructive feedforward ART clustering networks. I , 2002, IEEE Trans. Neural Networks.

[13]  IEEE-INNS-ENNS International Joint Conference on Neural Networks , 2000, Cognitive Systems Research.

[14]  R. Sabourin,et al.  Factors of overtraining with fuzzy ARTMAP neural networks , 2005, Proceedings. 2005 IEEE International Joint Conference on Neural Networks, 2005..

[15]  Georgios C. Anagnostopoulos,et al.  Ellipsoid ART and ARTMAP for incremental clustering and classification , 2001, IJCNN'01. International Joint Conference on Neural Networks. Proceedings (Cat. No.01CH37222).

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

[17]  Georgios C. Anagnostopoulos,et al.  Reducing generalization error and category proliferation in ellipsoid ARTMAP via tunable misclassification error tolerance: boosted ellipsoid ARTMAP , 2002, Proceedings of the 2002 International Joint Conference on Neural Networks. IJCNN'02 (Cat. No.02CH37290).

[18]  S. Grossberg,et al.  A self-organizing neural network for supervised learning, recognition, and prediction , 1992, IEEE Communications Magazine.