使用 tidymodels 调整工作流程集时如何正确设置参数网格？

Question

我尝试使用 tidymodels 通过配方和模型参数来调整工作流程。调整单个工作流程时没有问题。但是，当调整具有多个工作流程的工作流程集时，它总是会失败。这是我的代码：

\n

# read the training data\ntrain <- read_csv("../../train.csv")\ntrain <- train %>% \n    mutate(\n      id = row_number(),\n      across(where(is.double), as.integer),\n      across(where(is.character), as.factor),\n      r_yn = fct_relevel(r_yn, "yes")) %>% \n  select(id, r_yn, everything())\n\n# setting the recipes\n\n# no precess\nrec_no <- recipe(r_yn ~ ., data = train) %>%\n  update_role(id, new_role = "ID")\n\n# downsample: tuning the under_ratio\nrec_ds_tune <- rec_no %>% \n  step_downsample(r_yn, under_ratio = tune(), skip = TRUE, seed = 100) %>%\n  step_nzv(all_predictors(), freq_cut = 100)\n\n# setting the models\n\n# randomforest\nspec_rf_tune <- rand_forest(trees = 100, mtry = tune(), min_n = tune()) %>%\n  set_engine("ranger", seed = 100) %>%\n  set_mode("classification")\n\n# xgboost\nspec_xgb_tune <- boost_tree(trees = 100, mtry = tune(), tree_depth = tune(), learn_rate = tune(), min_n = tune()) %>% \n   set_engine("xgboost") %>% \n   set_mode("classification")\n\n# setting the workflowsets\nwf_tune_list <- workflow_set(\n  preproc = list(no = rec_no, ds = rec_ds_tune),\n  models = list(rf = spec_rf_tune, xgb = spec_xgb_tune),\n  cross = TRUE)\n\n# finalize the parameters, I\'m not sure it is correct or not\nrf_params <- spec_rf_tune %>% parameters() %>% update(mtry = mtry(c(1, 15)))\nxgb_params <- spec_xgb_tune %>% parameters() %>% update(mtry = mtry(c(1, 15)))\nds_params <- rec_ds_tune %>% parameters() %>% update(under_ratio = under_ratio(c(1, 5)))\n\nwf_tune_list_finalize <- wf_tune_list %>% \n  option_add(param = ds_params, id = c("ds_rf", "ds_xgb")) %>% \n  option_add(param = rf_params, id = c("no_rf", "ds_rf")) %>% \n  option_add(param = xgb_params, id = c("no_xgb", "ds_xgb"))\n

\n

我检查了wf_tune_list_finalize中的选项，它显示：

\n

> wf_tune_list_finalize$option\n[[1]]\na list of options with names:  \'param\'\n\n[[2]]\na list of options with names:  \'param\'\n\n[[3]]\na list of options with names:  \'param\'\n\n[[4]]\na list of options with names:  \'param\'\n

\n

然后我调整这个工作流程集：

\n

# tuning the workflowset\ncl <- makeCluster(detectCores())\nregisterDoParallel(cl)\nwf_tune_race <- wf_tune_list_finalize %>%\n  workflow_map(fn = "tune_race_anova",\n               seed = 100,\n               resamples = cv_5,\n               grid = 3,\n               metrics = metric_auc,\n               control = control_race(parallel_over = "everything"), \n               verbose = TRUE)\nstopCluster(cl)\n

\n

详细消息表明工作流程ds_rf和ds_xgb中的参数有问题：

\n

i 1 of 4 tuning:     no_rf\ni Creating pre-processing data to finalize unknown parameter: mtry\n\xef\xbf\xbd\xef\xbf\xbd 1 of 4 tuning:     no_rf (1m 44.4s)\ni 2 of 4 tuning:     no_xgb\ni Creating pre-processing data to finalize unknown parameter: mtry\n\xef\xbf\xbd\xef\xbf\xbd 2 of 4 tuning:     no_xgb (28.9s)\ni 3 of 4 tuning:     ds_rf\nx 3 of 4 tuning:     ds_rf failed with: Some tuning parameters require finalization but there are recipe parameters that require tuning. Please use `parameters()` to finalize the parameter ranges.\ni 4 of 4 tuning:     ds_xgb\nx 4 of 4 tuning:     ds_xgb failed with: Some tuning parameters require finalization but there are recipe parameters that require tuning. Please use `parameters()` to finalize the parameter ranges.\n

\n

结果是：

\n

> wf_tune_race\n# A workflow set/tibble: 4 x 4\n  wflow_id info             option      result        \n  <chr>    <list>           <list>      <list>        \n1 no_rf    <tibble [1 x 4]> <wrkflw__ > <race[+]>     \n2 no_xgb   <tibble [1 x 4]> <wrkflw__ > <race[+]>     \n3 ds_rf    <tibble [1 x 4]> <wrkflw__ > <try-errr [1]>\n4 ds_xgb   <tibble [1 x 4]> <wrkflw__ > <try-errr [1]>\n

\n

另外，虽然no_rf和no_xgb都有调优结果，但我发现这两个工作流程中mtry的范围不是我上面设置的范围，这意味着参数范围设置步骤完全失败。我已经按照https://www.tmwr.org/workflow-sets.html和https://workflowsets.tidymodels.org/的教程进行操作，但仍然没有任何想法。

\n

那么在调整工作流程集时如何正确设置配方和模型参数呢？

\n

我的代码中的train.csv在这里：https ://github.com/liuyifeikim/Some-data

\n

Answer 1

我修改了参数设置步骤，现在调整结果是正确的：

# setting the parameters on each workflow seperately
no_rf_params <- wf_set_tune_list %>% 
  extract_workflow("no_rf") %>% 
  parameters() %>% 
  update(mtry = mtry(c(1, 15)))

no_xgb_params <- wf_set_tune_list %>% 
  extract_workflow("no_xgb") %>% 
  parameters() %>% 
  update(mtry = mtry(c(1, 15)))

ds_rf_params <- wf_set_tune_list %>% 
  extract_workflow("ds_rf") %>% 
  parameters() %>% 
  update(mtry = mtry(c(1, 15)), under_ratio = under_ratio(c(1, 5)))

ds_xgb_params <- wf_set_tune_list %>% 
  extract_workflow("ds_xgb") %>% 
  parameters() %>% 
  update(mtry = mtry(c(1, 15)), under_ratio = under_ratio(c(1, 5)))

# update the workflowset
wf_set_tune_list_finalize <- wf_set_tune_list %>% 
  option_add(param_info = no_rf_params, id = "no_rf") %>%
  option_add(param_info = no_xgb_params, id = "no_xgb") %>% 
  option_add(param_info = ds_rf_params, id = "ds_rf") %>% 
  option_add(param_info = ds_xgb_params, id = "ds_xgb")

其余部分保持不变。我认为可能有一些有效的方法来设置参数。