Figure 1

Illustration of S3D learning a model of data. S3D bins the values of each feature (left panel) to select one that explains the largest variation of the outcome Y (middle panel). Next, it bins each unselected feature to identify one that explains the largest amount of the remaining variation. The aerial view of the partition of the important feature space (right panel) represents a visualization of the data, which is also represented by a tree (bottom panel). S3D continues partition the space using features not previously selected, until there is no appreciable variation left to explain