Table 5 |
||||||
|
The cross-validation performance of the NN method using various numbers of hidden neurons. |
||||||
|
Hidden Neurons |
H |
E |
O |
|||
|
|
||||||
|
δ |
r |
δ |
r |
δ |
r |
|
|
|
||||||
|
1 |
0.058 |
0.965 |
0.081 |
0.873 |
0.086 |
0.609 |
|
3 |
0.055 |
0.968 |
0.067 |
0.912 |
0.063 |
0.816 |
|
5 |
0.055 |
0.968 |
0.068 |
0.909 |
0.063 |
0.815 |
|
7 |
0.055 |
0.968 |
0.067 |
0.912 |
0.062 |
0.816 |
|
9 |
0.055 |
0.967 |
0.067 |
0.912 |
0.063 |
0.815 |
|
|
||||||
|
The secondary structure assignment scheme is the three-state α-helix (H), β-sheet (E) and other (O). The best results for each secondary structural type are highlighted in bold; they indicate that 7 neurons are marginally optimal overall for the SP175 dataset. |
||||||
|
Lees et al. BMC Bioinformatics 2006 7:507 doi:10.1186/1471-2105-7-507 |
||||||
