Cushion Treemaps and Botanical Visualization Yimeng Dou 05

Cushion Treemaps and Botanical Visualization Yimeng Dou 05 -16 -2002 ydou@ics. uci. edu

Overview q. Trees and Cushion Treemaps q. Sequoia. View Software q. Botanical Trees

Cushion Treemaps q. Provide shading as a strong extra cue to emphasize hierarchical structure. q. Especially good for answering global questions like “Why is my disk full? ”

Example: Trees

Example: Trees root

A disk with color scheme option off.

Same disk with color scheme on. Treemap

Treemap q. Efficient use of display space q. Constructed via recursive subdivision of the initial rectangle q. Size of each rectangle is proportional to the size of the node q. Subdivision is alternated per level: first horizontally, next vertically…

Treemap q Most useful when the feature we want to display is size q Not very good for visualizing structure of the tree (Worst case is balanced tree) q What happens if it’s a perfectly balanced tree of items all the same size?

Example: 3600 Employees

Nested Treemap q Use slightly smaller rectangles. Siblings are enclosed by a margin. q Require viewer’s effort when tree is deeply nested. q Coloring won’t help much. It does not provide a natural hierarchical structure, and we want to use color to show other attributes.

Nested Tree-Map Non-nested Tree-Map

The Idea Behind Cushion Treemaps q Human visual system interprets variations in shade as illuminated surfaces. So shape can be constructed to encode tree structure. Add a bump to each of the two subdivisions, and repeat recursively.

The Parabolic Surfice Z is height of such a surface.

Interaction—Sequoia. View q Sequoia. View is an interactive system for the analysis and visualization of large tree structures. q Cushion tree can be easily constructed and If there is any change, Sequoia can easily refresh the tree. q User can define size range, color scheme, can zoom in on sub-trees, zoon out again, and select preferred color scheme and filters.

<- Sound Garden Cure-Disintegration Sound Garden Studio Albums Only

Cushion Treemaps (Conclusion) q Efficient, quick generation of treemap image q Effective (shading provide a strong cue for identifying substructures) q Compact, no scrolling neccessary q Easy to implement (with the shown algorithm) q Easy to control with intuitive parameters q Wide applicability

Botanical Visualization of Huge Hierarchies—Idea q People can easily see branches, leaves and their arrangement in a botanical tree. q We can use the idea of botanical modeling for Information Visualization, and map folders to branches and files to leaves. q The model used—strand model (dates back to Leonardo Da Vinci) by Holton.

Strand Model • Leaves are connected to root via a strand • Area of branch is proportional to number of leaves.

Node and Link Diagram and Its Corresponding Botanical Tree

Problems q. Continuing branches representing a directory can’t be easily followed at the branching point. q. Those directories with many sub-directories lead to thin and long branches. q. Leaves tend to clutter.

An Example Of A Messy Tree

Continuation Without Extrusion • Smooth continuing branches (by adding a smooth transition between two cylinders). • It makes clear what the status of each branch is. • We can also use different color as an extra cue.

Contraction Long Branches • Conditionally remove the stem in the subbranch of the continuing branch. It replaces the binary tree with a general tree.

Files As Fruits q To prevent cluttering of leaves, we can use an icon to represent a list of files and their sizes. It can be modeled as a fruit consisting of a sphere with spots for each file. (Phi-ball) q Area of slices on the sphere is proportional to the size of the corresponding files. q When there is only one file, using cone instead of planar disks pasted on the sphere.

Phi-balls Cone’s length c equals to the square root of the ratio of the file size and the total size of the file list.

Final Results

Complete Hard Disk

A Unix Home Directory

Conclusions q. Cone covered phi-ball is good for visualizing a list of items, also useful for other applications q. Branches and cones hardly ever collide with no special prevention q. Efficient use of space by mathematical, algorithmic and physically based methods.

END Special thanks to Daniel Loewus. Deitch for providing graphs of his music library.