SeeGraph: A System for Visualizing Weighted-Edge Graphs

1
SeeGraphA System for Visualizing Weighted-Edge
Graphs

• Joshua New
• Dr. Jian Huang
• Dr. Elissa Chesler

2
Overview

• System Diagram
• Algorithmic Tools
• Screenshot

3
System Diagram
4
System Diagram

• .wel Weighted Edge List correlations with
  format
• NumVerts lttabgt NumEdges ltnewlinegt
• Vertex1 lttabgt Vertex2 lttabgt WeightV1toV2
  ltnewlinegt...
• .gvl Graph Vertex Layout - contains a 3D
  position (as a short int -32k,32k) in format
• Comment line ltnewlinegt NumVerts lttabgt NumEdges
  ltnewlinegt
• Vertex1 lttabgt V1_X lttabgt V1_Y lttabgt V1_Z
  ltnewlinegt
• .btv Block Tridiagonalization Vector BTDs
  permutation vector with NumVerts tab-delimited
  integers of row/col indices

5
System Diagram

• .txt Database Info - features file with
  format
• NumRows lttabgt NumCols ltnewlinegt
• Row1Col1 lttabgt Row1Col2 lttabgt... Row1ColN
  ltnewlinegt
• .map Database Index maps vertex names in the
  .txt file to data in .wel file with format
• VertexI ltnewlinegt VertexJ ltnewlinegt...
• _labels.map Vertex Names used to display on
  the graph when desired with format
• VertexI_DisplayName ltnewlinegt
• VertexJ_DisplayName ltnewlinegt ...

6
Algorithmic Tools

• Graph Layout O(MV2)

7
Edge Coloring
H o t / C o l d
NS0-.5,.5-1
N!S0-1
!NS0-.15,.85-1
!N!S.85-1
F u l l C o l o r
8
Dynamic Threshold
P A R A M 2
0.85 - 0.91
0.92 - 0.95
0.85 - 0.91
P A R A M 3
0.92 - 0.98
9
Algorithmic Tools

• Block tridiagonalization (BTD)
• Eigenvalues of the final matrix differ from those
  of the original matrix within a specified error
  tolerance
• The final matrix has small relative bandwidth
• Any offdiagonal blocks in the final matrix have
  either low dimension or are close to a low-rank
  matrix

10
Algorithmic Tools

• Adjacency Matrix and BTD Visualization

11
Algorithmic Tools

• Database query (quantitative selection)
• Neural Network (qualitative selection)

12
Algorithmic Tools

• Context
• Graph algorithms (induced subgraph, LCC, etc.)

13
Screenshot