Click on the figure to download the high definition version.
Figure 1
Figure 2
Figure 3
Figure 4
Left: Partial orders of the maximal segments. Middle: a layout order-equivalent to the left figure, changed maximal segments highlighted in green. Right: a layout not order equivalent to the other two figures. Red/blue arrows: relations between maximal segments.
TreemapsT′(with gray rectangle inserted),T, and T∗(with gray area spread over maximal segments).
Scatter plot of the average layout change between T and T′or T∗for a random 25% sample of all algorithms and datasets
Distribution of datasets over classes.
Figure 5
Figure 6
Figure 7
Figure 8
For each data class with at least 50 datasets, the ratioof the consistency score (visual quality on the left, stability on the right) between the data class and the baseline.
Visual quality: matrix plots for each data class with at least 50 datasets plus baseline (left top). In each matrix plot, rows correspond to algorithms, columns to datasets. The lighter the color, the better the relative performance, capped at 1 (purple).
Stability: matrix plots each data class with at least 50 datasets plus baseline (left top). In each matrix plot, rows correspond to algorithms, columns to datasets. The lighter the color, the better the relative performance, capped at 1 (purple).
Visual quality vs stability as function of the levels of hierarchy feature.
Figure 9
Figure 10
Figure 11
Figure 12
Visual quality vs stability as function of the variance of node weights feature.
Visual quality vs stability as function of the speed of weight change feature.
Visual quality vs stability as function of the insertions and deletions feature.
Relative ranking of treemapping algorithms for all data classes. Each table cell shows algorithms in top-down decreasing order of average visual quality (left column) and average stability (right column).
