Abstract-To identify the important attributes of complex system, which is high-dimensional and contain both discrete and continuous variables, this paper proposes a sensitivity analysis method of hierarchical hybrid fuzzy -neural network. We derive the sensitivity indexes of discrete and continuous variables through the differential method. To verify the effectiveness of our method, this study employed a man-made example and a remote sensing image classification example to test the performance of our method. The results show that our method can really identify the important variables of complex system and discover the relations between input and output variables; therefore, they can be applied to simplify the model and improve the classification accuracy of model.
In this paper, a new algorithm for the hierarchical hybrid fuzzy-neural network model is proposed. The Takagi-Sugeno model and triangular membership function are adopted in the fuzzy system, and the Lasso function of the coefficient contraction method is used to reduce the strong interaction among discrete input variables. In the end, an experimental test on surface features classification by using LANDSAT ETM+ remote sensing image data of Zhangping and Anxi in Fujian Province is conducted. Compared with other neural networks, the classification result with the proposed approach is the most accurate, which proves its feasibility and validity, and can be used as a new classification method for surface features on remote sensing images.
a new training algorithm for hierachical hybrid fuzzy -neural network (HHFNN) based on Takagi -Sugeno (T-S) fuzzy system is proposed in this paper. Triangular membership function is adopted. And to reduce the strong interaction among discrete input variables, coefficient contraction method is employed; ridge regression function is used in the THEN parts of fuzzy rules. At last, pyrimidines medical data is used in simulations; results show that our new algorithm gets an advantage in accuracy over the existing training algorithms for HHFNN and standard BP algorithm.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.