Sparse Graph network-of-networks Algorithms
The goal of our research is provide better solutions for the analyst to one of the most challenging problems in the intelligence analysis community: how to distill massive amounts of data into reduced forms that can feed into a decision matrix for presentation to an analyst or decision maker.
This unstructured data analysis can have applications with political coalitions and cultural LEGOs. Many other types of decision graphs can be applied in various areas. We will enable high-performance video analysis, which will prove useful in analysis of social networks.
This represents 512 YouTube videos, ~22.6 million frames, ~5.2 Tbytes. Each point is a video frame, each color is a different video, and coordinates are PCA projection of N-d feature vector into 3-D.
Hierarchical sparse graph analysis:
- Breadth vs. Depth
- Highest level: Decision graphs
- Lowest level: Lots of data (social networks)
- Intermediate levels: N degrees of separation
- Specific algorithms
- Sparse graph operators:
- Parallel prefix operators
- N-Dimensional Voronoi/Delaunay mesh generation
- Used to generate graph links for N-dimensional feature data
- Graph analysis:
- Minimal cycles (i.e., optimal sub-graphs)
- Graph partitioning
- Link analysis.