Abstract: The Gustafson-Kessel? fuzzy clustering algorithm is capable of detecting hyperellipsoidal clusters of different sizes and orientations by adjusting the covariance matrix of data, thus overcoming the drawbacks of conventional fuzzy c-means algorithm. In this paper, an adaptive version of the Gustafson-Kessel? algorithm is proposed. The way to adjust the covariance matrix iteratively is introduced by applying the Sherman-Morrison? matrix inversion procedure. The adaptive fuzzy clustering algorithm is implemented on the base of self-learning spiking neural network known as a realistic analog of biological neural systems that can perform fast data processing. Therefore, the proposed fuzzy spiking neural network that belongs to a new type of hybrid intelligent systems makes it possible both to perform fuzzy clustering tasks efficiently and to reduce data processing time considerably.

Keywords: computational intelligence, hybrid intelligent system, fuzzy clustering, adaptive Gustafson-Kessel? algorithm, self-learning spiking neural network, spiking neuron center, the temporal Hebbian learning.

ACM Classification Keywords: I.2.6 Artificial Intelligence: Learning – Connectionism and neural nets; I.5.1 Pattern Recognition: Models – Fuzzy set, Neural nets; I.5.3 Pattern Recognition: Clustering – Algorithms.

Link:

ADAPTIVE GUSTAFSON-KESSEL FUZZY CLUSTERING ALGORITHM BASED ON SELF-LEARNING SPIKING NEURAL NETWORK

Yevgeniy Bodyanskiy, Artem Dolotov, Iryna Pliss

http://www.foibg.com/ibs_isc/ibs-13/ibs-13-p02.pdf