- Moderator: uru
- Protocol: sor
- Start: 13:00
- End: 14:00
What should the graphs to be analyzed look like?
- Different node sizes (some nodes with high height, some not...)
- Multiple edges per node (incoming, outcoming)
- No hyper edges
- No ports, especially no NORTH/SOUTH ports on nodes (implicit we do have ports)
- No comment boxes
- Evenly spaced edge attachment points
Which sets of graphs should be analyzed?
- Randomized Ptolemy diagrams
- Randomized pseudo graphs (add some high degree nodes that have a higher height; node.height / degree(node) + 1 > spacing + average node size
- degree per side
Which analyses should be performed?
- Count number of axis-parallel edges
- Size of the graph drawing (height, width), height -> width?
- Sum up whitespace between nodes on a layer (respect the actual minimal spacing)
How should the analyses be presented?
- Table presenting the height of the graphs, and the number of axis-parallel edges where the node were placed with
- modified BK
- Compare these results to the results of the SimpleNodePlacer. The SimpleNodePlacer acts as a lower bound for a comparison of the height values.
- Compare these results to the results of optimization algorithm implemented by uru (ILP). This algorithm should act as a lower bound for a comparison of the number of axis-parallel edges.
- Think about appropriate graphical representations
- Think about statistical evaluation (mean, average, more sophisticated analyses, ...)