Skip to main content

Table 5 Multi-class remote fold recognition using the mismatch(5,2) kernel

From: Efficient use of unlabeled data for protein sequence classification: a comparative study

Method Error Top-5 Error Balanced Error Top-5 Balanced Error F1 Top-5 F1
Without clustering       
full seq. 50.16 21.82 67.17 32.55 37.43 71.40
region 42.83 13.68 61.43 22.63 40.36 79.19
no tails (full seq.) 50.16 21.82 71.81 32.59 30.17 69.12
max. length (full seq.) 52.44 24.43 77.31 39.17 23.98 65.22
With clustering       
full seq. 50.33 19.71 70.04 27.21 32.10 75.03
region 40.88 13.68 57.86 22.82 47.54 79.03
no tails (full seq.) 48.37 20.68 69.83 32.27 31.48 70.03
max. length (full seq.) 52.44 23.29 77.05 36.52 26.84 68.02
\