Table 2.
Success rates for each algorithm.
| Algorithm | Minimal | Simple | Medium | Full | Spanning | Europe | Robustness |
|---|---|---|---|---|---|---|---|
| Unweighted | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Weighted | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| RankEdge | 1.00 | 0.98 | 0.84 | 0.70 | 0.54 | 0.70 | 0.78 |
| MaxEdgeA | 1.00 | 0.78 | 0.74 | 0.48 | 0.36 | 0.62 | 0.63 |
| MaxEdgeB | 0.76 | 0.60 | 0.64 | 0.64 | 0.44 | 0.70 | 0.62 |
| MaxEdgeC | 1.00 | 1.00 | 0.84 | 0.48 | 0.44 | 0.48 | 0.66 |
| MaxWeighted | 1.00 | 0.88 | 0.72 | 0.63 | 0.38 | 0.68 | 0.68 |
| Deneubourg | 1.00 | 1.00 | 0.94 | 0.44 | 0.67 | 0.52 | 0.72 |
For each algorithm, we show the success rate on each simulated network. The last column summarizes the robustness of each method across all networks. Of the non-linear algorithms, RankEdge performs the best.