SeeGraph Results:

0) Graph Data
This small graph dataset was provided by Shawn Ericson and comes from an unknown source. An in-lab spring-embedded layout [54MB PPT] was computed on the undirected, weighted graph to create 3D coordinates for each vertex. The graph consists of 254 vertices and 401 edges each with weight greater than |0.5|. Edges are always colored with default hot/cold coloring scheme (0.5=blue, 1.0=red). Click on images for zip of hi-res screenshots.


I) Single-Query
Only one query starting with vertex of highest degree and slowly "growing the graph" by adding vertices of lower degree until the entire graph is shown.

Growing a graph

II) Single-Query 3D View
Query of vertices with degree one. Only singleton edges shown, connecting vertices that only connect to one another (thus not a connected component with the rest of the graph). Vertices without edges have degree one but connect to vertices with degree greater than one.

Rotation of 1-degree vertices

III) Multi-Query 3D View
3 queries used to color-code vertices according to their degree Red[1], Green[2,3], or Blue[4,21]. Note that query precedence diminishes as the query listing goes down (i.e. top query of [1,21] results in all red vertices).

Rotation Without Edges	With Edges

IV) Neural Net Filtering
Two clicks to select example and counter-example and neural net immediately color-codes the graph accordingly. We can remove edges to avoid clutter or simply filter out all counter-examples. This allows us to see only nodes similar to the one we left-clicked to color green initially. The edges are the induced subgraph showing that some of the items similar to what we wanted are correlated to one another.

Select Node(s) of Interest	Color-code Vertices Accordingly
Show without Edges	3D view of similar items

SeeGraph Requested Results: [.tar.gz (Ib-III)]

Ia) Images [tar.gz]
All PPM images at 2-degree rotations.

Ib) Camera Positions [txt]
Camera position format: angle coi.x coi.y coi.z eye.x eye.y eye.z up.x up.y up.z
The data from demo1.gvl (IIa) is normalized to 100³ so data is centered at (50,50,50) and is rotated about OpenGL's y-axis so up is (0,1,0).

IIa) Geometry [gvl]
"Graph Vertex Layout" format:

WeightedEdgeData.wel Threshold AbsoluteValueFlag AttractionFlag ForceGravity SpringLayoutAlgorithm
NumVerts NumEdges
VertexLabel X Y Z ...

Note: 2 lines of header data: parameters used to generate this layout make sure you load the same WeightedEdge and GraphVertexLayout files.

IIb) Geometry Read Routine [tar.gz]
Read and normalize routines for gvl data (VC6 workspace and makefile included).

III) Edge Weight Information [wel]
"Weighted-Edge List" format:

NumVerts NumEdges
Vertex1Label Vertex2Label WeightVertex1ToVertex2 ...

Note: Any ordering of weights can be supplied, stored internally as lower-triangular matrix.



FUTURE WORK
Email me if you have suggestions or things you'd like to see.