Training set size | F-measure | CI |
---|
1000 | 0.7175 | (0.7095, 0.7278) |
500 | 0.7167 | (0.7075, 0.7269) |
200 | 0.7131 (0.7108, 0.7140) | (0.7041, 0.7236) |
100 | 0.7061 (0.7050, 0.7127) | (0.6943, 0.7184) |
- Mean F-measure on S-STRAND2, with 95% confidence intervals shown in the last column. For the bottom two rows, training set size was sampled 11 times uniformly at random, and the median (20-, 80-percentiles) of the prediction accuracies from these samples are reported, along with confidence intervals for the medians.