Adding AIC/BIC information

[matzhaugen]

Is your feature request related to a problem? Please describe.
I'm not sure if this is 100% relevant to penalized cox regressions, but would it be useful to ad AIC and BIC metrics if cross-validation is inappropriate, e.g. due to data sparsity or for the reason of replicating other work?
In the same vein, a forward and backward selection method could also be employed (but that would probably go in another feature request).

Describe the solution you'd like
Add an AIC and BIC metric to the existing set of metrics.

Describe alternatives you've considered
As an alternative one could also compute the AIC/BIC oneself outside out this codebase.

References and existing implementations
An existing implementation of this can be found in the sciki learn package here

One reference paper that employs this with backward selection is here
https://doi.org/10.1016/j.jtcvs.2017.11.095

Again, not sure if the other criteria already implemented are perhaps more appropriate. If so, let me know.

[sebp]

I agree that having AIC/BIC for CoxnetSurvivalAnalysis would be nice. If you want to work on it, please feel free to open a pull request.

Regarding, forward/backward selection, sklearn's SelectKBest could be helpful in that regard, but as you said, this is best discussed in a separate issue.