Hello - Am I correct that the newly supported AFT model in xgboost only supports the watchlist/evals option, but not cross-validation? It seems like when I try to use the AFT objective with cross-validation (e.g. using the xgb.cv function in R), I get an error indicating the a ‘label’ is required - but for AFT what we actually pass is the label_lower_bound/label_upper_bound. Am I missing something? Is here a reason why CV is currently unsupported?
Thank you!
Survival Analysis with Accelerated Failure Time
Yes, CV is currently not supported for AFT. I (one of the original implementers) simply didn’t have a chance to implement it.