Fig. 2
From: A multiple kernel density clustering algorithm for incomplete datasets in bioinformatics
Illustration of potential cluster centroids automatically detected using the proposed MKDCI algorithm on the DLBCL-B dataset. a Scatter plot of the distribution of potential cluster centroids in which plots of delta(δ) against the rank of rho(ρ), and potential cluster centroids have significantly higher values of δ and ρ. b Scatter plot of the distribution of potential cluster centroids in which plots of delta(δ) against theta(θ) are generated with MKDCI’s automatic detection method, and potential cluster centroids are easier recognized in this region. c Scatter plot of the DLBCL-B dataset with three ground-truth clusters, ground-truth clusters are color labeled, and potential cluster centroids are labeled by its data point index with a square