Figure 1
From: Clustering gene expression data with a penalized graph-based metric
PKNNG's dependence on k : Results on artificial datasets. Evaluation of different number of neighbors (k) for the PKNNG metric. We use five artificial datasets that simulate gene expression problems and consider two clustering methods (PAM, top row, and HC, bottom row) and two base metrics (Euclidean, left column, and Pearson's correlation, right column).