R finetune tune_sim_anneal 通过模拟退火优化模型参数

tune_sim_anneal(object, ...)

# S3 method for model_spec
tune_sim_anneal(
  object,
  preprocessor,
  resamples,
  ...,
  iter = 10,
  param_info = NULL,
  metrics = NULL,
  initial = 1,
  control = control_sim_anneal()
)

# S3 method for workflow
tune_sim_anneal(
  object,
  resamples,
  ...,
  iter = 10,
  param_info = NULL,
  metrics = NULL,
  initial = 1,
  control = control_sim_anneal()
)

创建新设置

对于每个数字参数，可能值的范围以及任何转换都是已知的。当前值经过转换并缩放为 0 到 1 之间的值(基于可能的值范围)。生成位于rmin 和rmax 之间随机半径的球体上的一组候选值。删除不可行的值并随机选择一个值。该值将重新转换为原始单位和比例，并用作新设置。 control_sim_anneal() 的参数 radius 控制范围邻域大小。

对于分类参数和整数参数，每个参数都会以预定义的概率发生变化。 control_sim_anneal() 的 flip 参数可用于指定此概率。对于整数参数，使用附近的整数值。

当许多参数是非数字或唯一值很少的整数时，模拟退火搜索可能不是首选方法。在这些情况下，同一候选集可能会被测试多次。

冷却时间表

为了确定接受新值的概率，需要计算性能差异百分比。如果要最大化性能指标，则为 d = (new-old)/old*100 。概率计算为 p = exp(d * coef * iter)，而 coef 是用户定义的常量，可用于增加或减少概率。

control_sim_anneal() 的cooling_coef 可用于此目的。

终止标准

当找到新的全局最佳结果时，重新启动计数器会重置。

当找到新的全局最佳值或改进次优结果时，终止计数器会重置。

并行性

tune 和 finetune 包当前通过重新采样并行化。指定并行 back-end 将改善 sub-models 初始集(如果有)的生成。如果注册了并行后端，则搜索的每次迭代也会并行运行。

tune::tune_grid() , control_sim_anneal() , yardstick::metric_set()

# \donttest{
library(finetune)
library(rpart)
#> 
#> Attaching package: ‘rpart’
#> The following object is masked from ‘package:dials’:
#> 
#>     prune
library(dplyr)
#> 
#> Attaching package: ‘dplyr’
#> The following object is masked from ‘package:MASS’:
#> 
#>     select
#> The following objects are masked from ‘package:stats’:
#> 
#>     filter, lag
#> The following objects are masked from ‘package:base’:
#> 
#>     intersect, setdiff, setequal, union
library(tune)
library(rsample)
library(parsnip)
library(workflows)
library(ggplot2)

## -----------------------------------------------------------------------------
if (rlang::is_installed("modeldata")) {
  data(two_class_dat, package = "modeldata")

  set.seed(5046)
  bt <- bootstraps(two_class_dat, times = 5)

  ## -----------------------------------------------------------------------------

  cart_mod <-
    decision_tree(cost_complexity = tune(), min_n = tune()) %>%
    set_engine("rpart") %>%
    set_mode("classification")

  ## -----------------------------------------------------------------------------

  # For reproducibility, set the seed before running.
  set.seed(10)
  sa_search <-
    cart_mod %>%
    tune_sim_anneal(Class ~ ., resamples = bt, iter = 10)

  autoplot(sa_search, metric = "roc_auc", type = "parameters") +
    theme_bw()

  ## -----------------------------------------------------------------------------
  # More iterations. `initial` can be any other tune_* object or an integer
  # (for new values).

  set.seed(11)
  more_search <-
    cart_mod %>%
    tune_sim_anneal(Class ~ ., resamples = bt, iter = 10, initial = sa_search)

  autoplot(more_search, metric = "roc_auc", type = "performance") +
    theme_bw()
}
#> Optimizing roc_auc
#> Initial best: 0.81147
#> 1 ♥ new best           roc_auc=0.82651 (+/-0.004242)
#> 2 ♥ new best           roc_auc=0.83439 (+/-0.009983)
#> 3 ♥ new best           roc_auc=0.83912 (+/-0.007893)
#> 4 ♥ new best           roc_auc=0.84267 (+/-0.006609)
#> 5 ◯ accept suboptimal  roc_auc=0.83949 (+/-0.008792)
#> 6 ◯ accept suboptimal  roc_auc=0.83751 (+/-0.01065)
#> 7 + better suboptimal  roc_auc=0.84225 (+/-0.009044)
#> 8 ◯ accept suboptimal  roc_auc=0.83736 (+/-0.01063)
#> 9 + better suboptimal  roc_auc=0.84225 (+/-0.009044)
#> 10 ◯ accept suboptimal  roc_auc=0.84225 (+/-0.009044)
#> There were 10 previous iterations
#> Optimizing roc_auc
#> 10 ✔ initial            roc_auc=0.84267 (+/-0.006609)
#> 11 ◯ accept suboptimal  roc_auc=0.83949 (+/-0.008792)
#> 12 ◯ accept suboptimal  roc_auc=0.83736 (+/-0.01063)
#> 13 + better suboptimal  roc_auc=0.84225 (+/-0.009044)
#> 14 ◯ accept suboptimal  roc_auc=0.84225 (+/-0.009044)
#> 15 ◯ accept suboptimal  roc_auc=0.83751 (+/-0.01065)
#> 16 + better suboptimal  roc_auc=0.84225 (+/-0.009044)
#> 17 ◯ accept suboptimal  roc_auc=0.83949 (+/-0.008792)
#> 18 ✖ restart from best  roc_auc=0.8421 (+/-0.007206)
#> 19 ◯ accept suboptimal  roc_auc=0.83928 (+/-0.007538)
#> 20 ─ discard suboptimal roc_auc=0.82378 (+/-0.01017)

# }