Stragegies for High Connectedness
Posted: Tue Jun 22, 2010 3:52 pm
Highly connected graphs cannot be displayed as simple graphs abiding by some Graph Layout algorithm and still be visually useful.
There are three basic types of Logical Zoom developed so far:
Total, Restrictive, and Focused.
Total is the conventional approach, zooming the entire collection at once into the next level of detail. This has the disadvantage of complete unreadability of highly-connected graphs, but the advantage of straightforward simplicity of implementation and usage. Thus, this should be used when graphs are not so highly connected as to make this approach untenable.
Realizing that uniform total zoom is not workable with highly connected graphs, we must develop a new approach to logical zoom.
We should, first of all, lump nodes into collections such that there is a minimum of nodes in each collection. This should be the initial level of zoom.
Then, a user may select one of these collections and the detailed view of that collection will be displayed.
Restrictive zoom chooses not to visualize the connections between nodes in the selected collection and other collections. The only connections visualized are the connections intracollection edges. This keeps things minimal and simple, but restricts visibility of information about node connectedness.
What we would instead like to do, assuming the degree of connectedness is not altogether ridiculous, is preserve information about the extracollection connections of nodes within the selected collection, but not display all connections between all nodes. This is known as Focused Logical Zoom.
Some other considerations and potential implementation ideas:
Asymmetric focus: zoom into a selected collection, but don't zoom other collections; display extracollection connections as pointing to higher zoom level collection of target node.
Highlight edges of selected node. This could/should be used under any of the above implementations.
Comments and discussion welcome
There are three basic types of Logical Zoom developed so far:
Total, Restrictive, and Focused.
Total is the conventional approach, zooming the entire collection at once into the next level of detail. This has the disadvantage of complete unreadability of highly-connected graphs, but the advantage of straightforward simplicity of implementation and usage. Thus, this should be used when graphs are not so highly connected as to make this approach untenable.
Realizing that uniform total zoom is not workable with highly connected graphs, we must develop a new approach to logical zoom.
We should, first of all, lump nodes into collections such that there is a minimum of nodes in each collection. This should be the initial level of zoom.
Then, a user may select one of these collections and the detailed view of that collection will be displayed.
Restrictive zoom chooses not to visualize the connections between nodes in the selected collection and other collections. The only connections visualized are the connections intracollection edges. This keeps things minimal and simple, but restricts visibility of information about node connectedness.
What we would instead like to do, assuming the degree of connectedness is not altogether ridiculous, is preserve information about the extracollection connections of nodes within the selected collection, but not display all connections between all nodes. This is known as Focused Logical Zoom.
Some other considerations and potential implementation ideas:
Asymmetric focus: zoom into a selected collection, but don't zoom other collections; display extracollection connections as pointing to higher zoom level collection of target node.
Highlight edges of selected node. This could/should be used under any of the above implementations.
Comments and discussion welcome