Specifies the name of the executable file within the Contents/MacOS directory. Kernels indicate more directly which points should be kept together, without considering the reference of a cluster centre. In the kernel case, the point of view is different: there is no explicit representation of the cluster prototypes, and the comparison function involves pairs of data points, and not cluster centres. Yet the aim of the latter differs from that of kernel methods: they look for clusters whose prototypes have a different nature than the data point, as for instance lines or quadric curves therefore they require a modification of the distance between data points and cluster centres. It is to be noticed that there exist other approaches to replace the euclidian distance by more appropriate ones in the fuzzy c –means algorithm, as the fuzzy shell approach for instance. The centre expression obtained in the non-kernel case, it can be shown that the update equations for the linear combination coefficients and the membership degrees only depend on the scalar products and thus on the kernel. Provided the cluster centres are looked for as linear combinations of the data, which is consistent with It consists in transposing the cost function to the feature space, i.e. The kernel extension of the fuzzy c-means is less direct as this algorithm explicitely uses the data points themselves. Lesot, in Modern Information Processing, 2006 2.2.2 Fuzzy c-means with kernels Kernel-based outlier preserving clustering with representativity coefficients

Where is the kernel in mac disk image