mlr - 无法在 ranger_permutation 过滤器中使用参数“重要性”

Question

When I try to use the filter 'ranger_permutation' and pass some parameters to the filter in makeFilterWrapper, I find that I cannot use the parameter 'importance' as that name is clashing with another parameter, but I cannot understand where or why. I get the following error message when I run the code below:

Error in setHyperPars(learner = wl, ..., par.vals = par.vals): 
Assertion on 'parameter settings' failed: Must have unique names, but element 4 is duplicated

I get this error when using other base learners as well. If I remove the parameter 'importance', the problem goes away, but not if I remove other parameters. Also I can run ranger on its own, not as a filter, with the same parameters.

library(survival)
#> Warning: package 'survival' was built under R version 3.5.3
library(mlr)
#> Loading required package: ParamHelpers

data(veteran)
set.seed(24601)
task_id = "VET"
vet.task <- makeSurvTask(id = task_id, data = veteran, target = c("time", "status"))
vet.task <- createDummyFeatures(vet.task)
outer = makeResampleDesc("CV", iters=2, stratify=TRUE)


set.seed(24601, "L'Ecuyer") 
cox.lrn <- makeLearner(cl="surv.coxph", id = "Ranger_Cox", predict.type="response")
filt <- makeFilterWrapper(cox.lrn, fw.method="ranger_permutation", fw.abs=5, cache=TRUE, num.trees=1000, splitrule="maxstat", importance="permutation")
bmr = benchmark(filt, vet.task, outer, list(cindex), show.info = TRUE, models=TRUE, keep.extract=TRUE)      
#> Task: VET, Learner: Ranger_Cox.filtered
#> Resampling: cross-validation
#> Measures:             cindex
#> Error in setHyperPars(learner = wl, ..., par.vals = par.vals): Assertion on 'parameter settings' failed: Must have unique names, but element 4 is duplicated.

^{Created on 2019-09-25 by the reprex package (v0.3.0)}

Answer 1

When you set fw.method = "ranger_permutation" importance = "permutation" is already set internally. Therefore when you force importance = "permutation" it is defined two times (duplicated) and the error checking complains.

This can be seen from the mlr implementation: https://github.com/mlr-org/mlr/blob/master/R/Filter.R makeFilter(name = "ranger_permutation"... has importance = "permutation" already defined.