Table 7.
Classification Performance on the Local Structure Alphabet Dataset.
| w = f = 5 | w = f = 7 | w = f = 9 | ||||
|---|---|---|---|---|---|---|
| ROC | Q16 | ROC | Q16 | ROC | Q16 | |
 |
0.82 | 64.9 | 0.81 | 64.7 | 0.81 | 64.2 |
 |
0.83 | 67.3 | 0.82 | 67.7 | 0.82 | 67.7 |
 |
0.84 | 66.4 | 0.84 | 66.9 | 0.83 | 67.2 |
 |
0.85 | 68.0 | 0.84 | 68.5 | 0.83 | 68.9** |
w = f gave the best results on testing on few sample points, and hence due to the expensive nature of this problem, we did not test it on a wide set of parameters. ** denotes the best scoring model based on the Q16 scores. For this best model the se rate of 0.21.