Table 1 Best learning subset: accuracy with different m and α in the PTB database
m\α | 1 | 2 | 4 | 8 | 16 | 32 |
|---|---|---|---|---|---|---|
4 | 68.45% | 66.99% | 75.73% | 72.33% | 67.96% | 67.96% |
5 | 65.53% | 67.96% | 68.45% | 70.39% | 67.96% | 67.96% |
6 | 66.99% | 72.33% | 71.36% | 69.90% | 68.93% | 65.05% |
7 | 68.93% | 69.42% | 75.24% | 73.30% | 68.45% | 68.45% |
8 | 69.42% | 67.96% | 76.70% | 80.58% | 67.48% | 67.96% |
9 | 66.99% | 66.50% | 73.79% | 78.16% | 71.84% | 70.39% |
10 | 68.93% | 78.64% | 80.10% | 77.18% | 69.90% | 68.45% |
11 | 72.33% | 77.67% | 75.24% | 66.02% | 68.93% | 70.39% |