For example, when there are lots of redundant or strongly
correlated attributes, the number of different trees can be
exponential to the number of related attributes. Digit recognition
data set [10] is such an example. All seven attributes
(horizontal and vertical lights in on-off combinations) are
strongly related. In this case, we need alternative tools to
help user understand why the tree structure might be so unstable.
Pixel bar charts technique could be such a tool [20]