如何将精选的功能列表添加到 mlr3 管道中过滤器的结果中

Question

我需要将选定数量的预选功能添加到 mlr3 管道中。我希望我可以将这些功能添加到另一个过滤器的结果中，如 mlr3 图表的这个片段所示：

这可以通过自定义过滤器或其他方法实现吗？

我浏览了 mlr3 书，没有找到任何解决这个问题的内容。

#simple dummy example of what I'm trying to achieve
library(mlr3filters)
library(mlr3fselect)
task_pen = tsk("penguins")
#feature selection using information gain
po_flt = po("filter", filter = flt("information_gain"), filter.nfeat = 3)
graph = po_flt
result = graph$train(task_pen)
result$featureunion.output$feature_names
#the output:
#"bill_depth"     "bill_length"    "flipper_length"

#However, I want to be able to add a list of curated features to the filter
po_custom = #a custom filter that takes a list of features,
#in this case "island" and "sex"
po_custom$id <- "curated list"
graph = gunion(list(
    po_flt,
    po_custom
  )) %>>%
  po("featureunion")
result = graph$train(task_pen)
result$featureunion.output$feature_names
#the output (obviously not runable):
#"bill_depth"     "bill_length"    "flipper_length"    "island"    "sex"

Answer 1

我相信我已经弄清楚了。 mlr3 中的选择器功能可以实现此功能（以及更多功能）：

po_flt = po("filter", filter = flt("information_gain"), filter.nfeat = 3)
pos = po("select", selector = selector_name(c("sex", "body_mass")))
graph = gunion(list(
    po_flt,
    pos
  )) %>>%
  po("featureunion")
result = graph$train(task_pen)
result$featureunion.output$feature_names
#"bill_depth"     "bill_length"    "flipper_length"    "body_mass"     "sex"