Fig. 2

Unboxing Approach. Illustrates how the data matrix is “unboxed” and embedded into the hierarchical clustering of a symmetric matrix. The process is similar for asymmetric matrices, except there are no redundant cells to remove. Top Left: A standard cluster heatmap of a correlation matrix. Top Middle: The cluster heatmap with non-redundant information highlighted. Top Right: The cluster heatmap without the redundant information. Bottom Left: Hierarchical clustering of the variables (rows and columns) from the cluster heatmap, shown as a dendrogram. Bottom Middle: Hierarchical clustering of values (individual cells) and variables, shown as a dendrogram. Bottom Right: Hierarchical clustering of values and variables, shown as a treemap