如何在 mlr3 中调整两个值的向量

Answer 1

您需要分别调整向量的各个部分，然后在转换中组合。您可能会发现 this tutorial 很有帮助。下面的例子

library(mlr3)
library(mlr3proba)
library(mlr3tuning)
library(mlr3extralearners)

t = tgen("simsurv")$generate(5)

search_space = ps(
  gamma = p_dbl(1e-3, 1e3),
  mu = p_dbl(1e-3, 1e3)
)
search_space$trafo = function(x, param_set) {
  x$gamma.mu = c(x$gamma, x$mu)
  x$gamma = x$mu = NULL
  x
}

AutoTuner$new(
  lrn("surv.svm", type = "hybrid", diff.meth = "makediff3",
        gamma.mu = c(0.1, 0.1)),
  rsmp("holdout"),
  msr("surv.cindex"),
  trm("evals", n_evals = 3),
  tnr("grid_search", resolution = 2),
  search_space
)$train(t)$predict(t)
#> INFO  [21:40:21.261] [bbotk] Starting to optimize 2 parameter(s) with '<OptimizerGridSearch>' and '<TerminatorEvals> [n_evals=3]' 
#> INFO  [21:40:21.275] [bbotk] Evaluating 1 configuration(s) 
#> INFO  [21:40:21.294] [mlr3] Running benchmark with 1 resampling iterations 
#> INFO  [21:40:21.315] [mlr3] Applying learner 'surv.svm' on task 'simsurv' 
#> INFO  [21:40:21.425] [bbotk] Finished optimizing after 3 evaluation(s) 
#> INFO  [21:40:21.425] [bbotk] Result: 
#> INFO  [21:40:21.426] [bbotk]  gamma   mu learner_param_vals  x_domain surv.cindex 
#> INFO  [21:40:21.426] [bbotk]  0.001 1000          <list[3]> <list[1]>           1
#> <PredictionSurv> for 5 observations:
#>  row_ids     time status      crank  response
#>        1 5.000000  FALSE -22.915171 22.915171
#>        2 5.000000  FALSE  -1.442179  1.442179
#>        3 3.640326   TRUE   3.849970 -3.849970
#>        4 0.268220   TRUE   4.269720 -4.269720
#>        5 5.000000  FALSE  -2.670886  2.670886

^{由reprex package (v2.0.1) 于 2022 年 2 月 19 日创建}