Kernel Sammon Map
Fernando Kentaro Inaba, Evandro Ottoni Teatini Salles, Thomas Walter Rauber · 2011
We extend the visualization technique of high-dimensional patterns conceived by Sammon to the case when the patterns have been previously mapped to an implicitly defined Hilbert feature space in which distances can be measured by kernels. The principal benefit of our technique is the possibility to gain insight into the distribution of the patterns, even in this generally non-accessible feature space.