Table 5.
Comparison of computation times (in seconds) taken for training coxnet and our method
LD pediatric trauma |
HD gene expression |
|||
---|---|---|---|---|
Model type | L2 | L1 | L2 | L1 |
coxnet | ![]() |
![]() |
![]() |
![]() |
Our method | ![]() |
![]() |
![]() |
![]() |
coxnet | ![]() |
![]() |
![]() |
![]() |
Our method | ![]() |
![]() |
![]() |
![]() |
Top: the cross-validation for both the methods was performed by varying the hyperparameter value between 10−3 and 109 by multiples of 10. Bottom: the cross-validation for both the methods was performed over the regularization path chosen by coxnet. The corresponding penalized log-likelihood achieved at convergence are in parentheses. Bold emphases represent superior performance of one method over the other.