A Learning Pattern Recognition Algorithm Using Fuzzy

Self-Organizing Map Network

Suliman M. Mohamed1, Henry O. Nyongesa1 and El-Sayed M. El-Horbaty2

(1)Dept. of Information Systems and Computing, Brunel University, Uxbridge, Middx UB8 3PH, UK

(2)Dept. of Computer Science, Faculty of Computer & Information Sciences, Ain Shams University, Cairo, EGYPT

Abstract:
This paper presents a learning pattern recognition algorithm, which uses synthetic technology of fuzzy self-organizing map network (FSOMIN). Our algorithm (fuzzy self-organizing map learning (FSOML)) depends on integrating unsupervised neural network namely Kohonen's self-organizing feature map (KSOFM) with fuzzy logic. Unsupervised classification algorithms are often based on a concept called data clustering and feature abstraction. KSOFMs are fuzzified to form a class of FSOMN. Unlike the crisp self-organizing feature map learning (SOFML) algorithm, were only one neuron will win and learn at each competition, in the FSOMNs every neuron will win at certain degree depending on its distance to the input pattern, and learn the pattern accordingly. Thus, the concept of win has been formulated as a fuzzy set and the network outputs become the win memberships in {0.1} of the competing neurons. Compared with crisp self-organizing map learning (SOML) algorithm, the FSOML algorithm converging, more often to the desired solutions, reducing the likelihood of neuron under-utilization performance, especially in overlapping data sets.

Keywords: Image Processing, Fuzzy Logic, Fuzzy-Self Organizing Map, Neural
                   Networks