Is there a way to use surv.aorsf with Harrel's concordance in MLR3?

Question

I am trying to use the surv.aorsf implementation in mlr3extralearners with the ’cstat’ split_rule setting. This parameter is valid in the orsf package. on the surv.aorsf mlr3 page (see here) the parameter is not listed, but in the example code at the bottom it is included.

When I run that code:

learner = mlr3::lrn("surv.aorsf")
learner$param_set$ids()

the parameter is not included, and when I try to initialise a learner with the parameter specified, i.e. with

ao_tuning_L = lrn("surv.aorsf",split_rule="cstat",
                   split_min_events = to_tune(2,10),
                   mtry = to_tune(1,9),
                   n_retry = to_tune(0, 5),
                   n_tree = to_tune(100, 1000))

I get the following error: "Error in x$public_methods : object of type 'closure' is not subsettable" This works for other models, such as surv.cforest and the "teststat" parameter.

I wonder if the parameter was omitted for a reason or overlooked? In order to do this I think I will need to build a custom learner, but I am not sure where to start. Is there a way to import all of the settings for surv.aorsf except for the split_rule?

--------- EDIT:

with updated mlr3extralearners and mlr3tuning, when running the next bit of code:

tuner_resampling =2
runs=2
rats<-survival::rats
rats$sex<-(rats$sex=="f")*1
tsk_rats = as_task_surv(rats, time = "time",
                        event = "status", type = "right", id = "rats")

ao_tuning_L = lrn("surv.aorsf",split_rule="cstat")
ao_tuner_L = tnr("random_search")
set_threads(ao_tuning_L, n = threads)
ao_instance_L = ti(tsk_rats,
                   learner=ao_tuning_L,
                   resampling = rsmp("cv",folds=tuner_resampling),
                   measure = msr("surv.cindex"),
                   terminator = trm("evals", n_evals = runs, k = 0))
ao_hyper_L <- ao_tuner_L$optimize(ao_instance_L)

I get the following error: "Error: there were unrecognized arguments: split_rule is unrecognized - did you mean n_split?"

Answer 1

The parameter split_rule was added 3 days ago. You therefore need to update your version of mlr3extralearners by for example running:

remotes::install_github("mlr-org/mlr3extralearners")

PS: Updating a package to the latest dev version is almost always a good idea to troubleshoot issues.