The Research On RSOM-based Tree-Structured Hybrid Learning Model

Tree-structured hybrid learning models have been introduced by a number of authors in recent years. It combines the advantages of both symbolic and non-symbolic models to solve tough problems and indicates great potential application in the fields of pattern recognition, biomedicine, information security, fault diagnosis and facial expression analysis, etc. The research of this paper is based on an RSOM tree structured hybrid learning model.In this paper, the fundamental of neural net work tree model is introduced, and the general structure of the model is constructed. The RSOM tree, which is a kind of hybrid decision tree and in which each internal node is embedded with a modular SOM net, is also introduced. And an increment training method of the RSOM tree is presented so that the model is able to adapt to learning in the dynamic and incremental case. Considering that it is necessary to select the available features from all a large group of features, an iterative evolution training mehod of the classifier model by combining the RSOM tree with Tabu search algorithm and the genetic algorithm is researched. By which, the available features are selected in a dynamic process.Meanwhile, the RSOM recognition algorithm makes a comprehensive decision by averaging the sum of weights. This may recognize some new samples which are distinguishable from original training samples as one of the known classes in the leaf node, whose sample number is predominant. Obviously, it might lead to make wrong decisions. Therefore, we construct an RSOM-RBF tree and an RSOM-SVM tree. The experiments show that these hybrid learning models are feasible.