Skip to main content


Fig. 2 | Big Data Analytics

Fig. 2

From: Identification of disease-distinct complex biomarker patterns by means of unsupervised machine-learning using an interactive R toolbox (Umatrix)

Fig. 2

Scheme of the workflow of generating a U-matrix with subsequent identification of clusters in the data using the R package “Umatrix” ( The process starts with a topology-preserving projection of high-dimensional data points onto a two-dimensional self-organizing network consisting of a grid of neurons of the Kohonen type [13] obtained via training of an emergent self-organizing map (ESOM). Subsequently, the distances between data points are projected on top of the grid as so-called U-matrix. Since the ESOM is toroid, i.e., opposite edges are connected, a so-called island is cut off the U-matrix, which contains each part of the U-matrix only once. Finally, on the finalized U-matrix clusters can be obtained interactively

Back to article page