If only data-data links are enabled then data nodes will form into clusters of similar items (Figure 3). The form of clustering has strong similarities with k-nearest neighbour classification, because data-data links are formed using a simple, knowledge free, comparison of data items, and only the strongest K links are kept.
Figure 3: Soybean dataset clustered by data-data links
Inter-cluster relationships can also be seen, where some nodes in one cluster have links with nodes in another. These links arrange the clusters so that the relationships between clusters are visible.