A hierarchical fuzzy clustering framework for training RBF networks
Proceedings of the IASTED International Conference on Artificial Intelligence and Soft Computing, ASC 2011, 2011
ABSTRACT This paper proposes a new method that combines input-output fuzzy clustering and optimal... more ABSTRACT This paper proposes a new method that combines input-output fuzzy clustering and optimal fuzzy clustering for the efficient design of radial basis function neural networks. We first apply the fuzzy c-means in the product (i.e. input-output) space to pre-process the available data. The resulting clusters are projected on the input space. The corresponding cluster centers are considered as a new data set which is further clustered by means of optimal fuzzy clustering in terms of the weighted fuzzy c-means. To accomplish this task we develop a new cluster validity index, which is used to identify the appropriate number of RBF hidden nodes. The algorithm is successfully implemented in well-known data sets where its performance is tested and evaluated.
Uploads
Papers by Mike Kenteris