A Survey on Kernelized Fuzzy Clustering Tools for Data Processing

Although most of the well-known fuzzy clustering algorithms are somewhat sensitive to noise, there are some more profound possibilistic fuzzy clustering techniques based on kernel distance metrics that are not subject to this problem. The paper presents a survey on this type of fuzzy clustering methods. Meanwhile, introducing a particular distance measure based on the Bregman divergence to a fuzzy clustering tool made it possible to improve the algorithm’s performance compared to traditional Euclidean-based analogs. A bunch of experimental evaluation is performed for this set of methods.