|
w
|
f= 1
|
f= 3
|
f= 5
|
f= 7
|
f= 9
|
f= 11
|
---|
| |
CC
|
RMSE
|
CC
|
RMSE
|
CC
|
RMSE
|
CC
|
RMSE
|
CC
|
RMSE
|
CC
|
RMSE
|
|
3
|
0.704
|
0.696
|
0.708
|
0.692
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
|
7
|
0.712
|
0.683
|
0.719
|
0.677
|
0.723
|
0.672
|
0.722
|
0.672
|
-
|
-
|
-
|
-
|
|
11
|
0.711
|
0.681
|
0.720
|
0.673
|
0.725
|
0.667
|
0.725
|
0.666
|
0.724
|
0.666
|
0.722
|
0.667
|
|
15
|
0.709
|
0.680
|
0.719
|
0.672
|
0.726**
|
0.665
|
0.726
|
0.664
|
0.725
|
0.664
|
0.723
|
0.664
|
- CC and RMSE denotes the average correlation coefficient and RMSE values. The numbers in bold show the best models as measured by CC for a fixed w parameter.
, and
represent the PSI-BLAST profile and YASSPP scoring matrices, respectively. soe, rbf, and lin represent the three different kernels studied using the
as the base kernel. * denotes the best regression results in the sub-tables, and ** denotes the best regression results achieved on this dataset. For the best results the se rate for the CC values is 0.003. The published results [15] uses the default rbf kernel to give CC = 0.600 and RMSE = 0.78.