# nolint start
library(mlexperiments)
library(mllrnrs)
# nolint start
library(mlexperiments)
library(mllrnrs)
See https://github.com/kapsner/mllrnrs/blob/main/R/learner_glmnet.R for implementation details.
library(mlbench)
data("DNA")
<- DNA |>
dataset ::as.data.table() |>
data.tablena.omit()
<- colnames(dataset)[160:180]
feature_cols <- "Class" target_col
<- 123
seed if (isTRUE(as.logical(Sys.getenv("_R_CHECK_LIMIT_CORES_")))) {
# on cran
<- 2L
ncores else {
} <- ifelse(
ncores test = parallel::detectCores() > 4,
yes = 4L,
no = ifelse(
test = parallel::detectCores() < 2L,
yes = 1L,
no = parallel::detectCores()
)
)
}options("mlexperiments.bayesian.max_init" = 10L)
<- splitTools::partition(
data_split y = dataset[, get(target_col)],
p = c(train = 0.7, test = 0.3),
type = "stratified",
seed = seed
)
<- model.matrix(
train_x ~ -1 + .,
$train, .SD, .SDcols = feature_cols]
dataset[data_split
)<- dataset[data_split$train, get(target_col)]
train_y
<- model.matrix(
test_x ~ -1 + .,
$test, .SD, .SDcols = feature_cols]
dataset[data_split
)<- dataset[data_split$test, get(target_col)] test_y
<- splitTools::create_folds(
fold_list y = train_y,
k = 3,
type = "stratified",
seed = seed
)
# required learner arguments, not optimized
<- list(
learner_args family = "multinomial",
type.measure = "class",
standardize = TRUE
)
# set arguments for predict function and performance metric,
# required for mlexperiments::MLCrossValidation and
# mlexperiments::MLNestedCV
<- list(type = "response", reshape = TRUE)
predict_args <- metric("bacc")
performance_metric <- NULL
performance_metric_args <- FALSE
return_models
# required for grid search and initialization of bayesian optimization
<- expand.grid(
parameter_grid alpha = seq(0, 1, 0.05)
)# reduce to a maximum of 10 rows
if (nrow(parameter_grid) > 10) {
set.seed(123)
<- sample(seq_len(nrow(parameter_grid)), 10, FALSE)
sample_rows <- kdry::mlh_subset(parameter_grid, sample_rows)
parameter_grid
}
# required for bayesian optimization
<- list(
parameter_bounds alpha = c(0., 1.)
)<- list(
optim_args iters.n = ncores,
kappa = 3.5,
acq = "ucb"
)
<- mlexperiments::MLTuneParameters$new(
tuner learner = mllrnrs::LearnerGlmnet$new(
metric_optimization_higher_better = FALSE
),strategy = "grid",
ncores = ncores,
seed = seed
)
$parameter_grid <- parameter_grid
tuner$learner_args <- learner_args
tuner$split_type <- "stratified"
tuner
$set_data(
tunerx = train_x,
y = train_y
)
<- tuner$execute(k = 3)
tuner_results_grid #>
#> Parameter settings [==================>-----------------------------------------------------------------------------] 2/10 ( 20%)
#> Parameter settings [============================>-------------------------------------------------------------------] 3/10 ( 30%)
#> Parameter settings [=====================================>----------------------------------------------------------] 4/10 ( 40%)
#> Parameter settings [===============================================>------------------------------------------------] 5/10 ( 50%)
#> Parameter settings [=========================================================>--------------------------------------] 6/10 ( 60%)
#> Parameter settings [==================================================================>-----------------------------] 7/10 ( 70%)
#> Parameter settings [============================================================================>-------------------] 8/10 ( 80%)
#> Parameter settings [=====================================================================================>----------] 9/10 ( 90%)
#> Parameter settings [===============================================================================================] 10/10 (100%)
head(tuner_results_grid)
#> setting_id metric_optim_mean lambda alpha family type.measure standardize
#> 1: 1 0.4728578 0.003092562 0.70 multinomial class TRUE
#> 2: 2 0.4737550 0.002639842 0.90 multinomial class TRUE
#> 3: 3 0.4733064 0.003330451 0.65 multinomial class TRUE
#> 4: 4 0.4733064 0.017972493 0.10 multinomial class TRUE
#> 5: 5 0.4733064 0.003993887 0.45 multinomial class TRUE
#> 6: 6 0.4728578 0.022574498 0.05 multinomial class TRUE
<- mlexperiments::MLTuneParameters$new(
tuner learner = mllrnrs::LearnerGlmnet$new(
metric_optimization_higher_better = FALSE
),strategy = "bayesian",
ncores = ncores,
seed = seed
)
$parameter_grid <- parameter_grid
tuner$parameter_bounds <- parameter_bounds
tuner
$learner_args <- learner_args
tuner$optim_args <- optim_args
tuner
$split_type <- "stratified"
tuner
$set_data(
tunerx = train_x,
y = train_y
)
<- tuner$execute(k = 3)
tuner_results_bayesian #>
#> Registering parallel backend using 4 cores.
head(tuner_results_bayesian)
#> Epoch setting_id alpha gpUtility acqOptimum inBounds Elapsed Score metric_optim_mean lambda errorMessage family
#> 1: 0 1 0.70 NA FALSE TRUE 1.666 -0.4728578 0.4728578 0.003092562 NA multinomial
#> 2: 0 2 0.90 NA FALSE TRUE 1.635 -0.4737550 0.4737550 0.002639842 NA multinomial
#> 3: 0 3 0.65 NA FALSE TRUE 1.657 -0.4733064 0.4733064 0.003330451 NA multinomial
#> 4: 0 4 0.10 NA FALSE TRUE 1.744 -0.4733064 0.4733064 0.017972493 NA multinomial
#> 5: 0 5 0.45 NA FALSE TRUE 0.577 -0.4733064 0.4733064 0.003993887 NA multinomial
#> 6: 0 6 0.05 NA FALSE TRUE 0.687 -0.4728578 0.4728578 0.022574498 NA multinomial
#> type.measure standardize
#> 1: class TRUE
#> 2: class TRUE
#> 3: class TRUE
#> 4: class TRUE
#> 5: class TRUE
#> 6: class TRUE
<- mlexperiments::MLCrossValidation$new(
validator learner = mllrnrs::LearnerGlmnet$new(
metric_optimization_higher_better = FALSE
),fold_list = fold_list,
ncores = ncores,
seed = seed
)
$learner_args <- tuner$results$best.setting[-1]
validator
$predict_args <- predict_args
validator$performance_metric <- performance_metric
validator$performance_metric_args <- performance_metric_args
validator$return_models <- return_models
validator
$set_data(
validatorx = train_x,
y = train_y
)
<- validator$execute()
validator_results #>
#> CV fold: Fold1
#>
#> CV fold: Fold2
#>
#> CV fold: Fold3
head(validator_results)
#> fold performance alpha lambda family type.measure standardize
#> 1: Fold1 0.3672581 0.7 0.003092562 multinomial class TRUE
#> 2: Fold2 0.3524351 0.7 0.003092562 multinomial class TRUE
#> 3: Fold3 0.3700659 0.7 0.003092562 multinomial class TRUE
<- mlexperiments::MLNestedCV$new(
validator learner = mllrnrs::LearnerGlmnet$new(
metric_optimization_higher_better = FALSE
),strategy = "grid",
fold_list = fold_list,
k_tuning = 3L,
ncores = ncores,
seed = seed
)
$parameter_grid <- parameter_grid
validator$learner_args <- learner_args
validator$split_type <- "stratified"
validator
$predict_args <- predict_args
validator$performance_metric <- performance_metric
validator$performance_metric_args <- performance_metric_args
validator$return_models <- return_models
validator
$set_data(
validatorx = train_x,
y = train_y
)
<- validator$execute()
validator_results #>
#> CV fold: Fold1
#>
#> Parameter settings [==================>-----------------------------------------------------------------------------] 2/10 ( 20%)
#> Parameter settings [============================>-------------------------------------------------------------------] 3/10 ( 30%)
#> Parameter settings [=====================================>----------------------------------------------------------] 4/10 ( 40%)
#> Parameter settings [===============================================>------------------------------------------------] 5/10 ( 50%)
#> Parameter settings [=========================================================>--------------------------------------] 6/10 ( 60%)
#> Parameter settings [==================================================================>-----------------------------] 7/10 ( 70%)
#> Parameter settings [============================================================================>-------------------] 8/10 ( 80%)
#> Parameter settings [=====================================================================================>----------] 9/10 ( 90%)
#> Parameter settings [===============================================================================================] 10/10 (100%)
#> CV fold: Fold2
#> CV progress [====================================================================>-----------------------------------] 2/3 ( 67%)
#>
#> Parameter settings [==================>-----------------------------------------------------------------------------] 2/10 ( 20%)
#> Parameter settings [============================>-------------------------------------------------------------------] 3/10 ( 30%)
#> Parameter settings [=====================================>----------------------------------------------------------] 4/10 ( 40%)
#> Parameter settings [===============================================>------------------------------------------------] 5/10 ( 50%)
#> Parameter settings [=========================================================>--------------------------------------] 6/10 ( 60%)
#> Parameter settings [==================================================================>-----------------------------] 7/10 ( 70%)
#> Parameter settings [============================================================================>-------------------] 8/10 ( 80%)
#> Parameter settings [=====================================================================================>----------] 9/10 ( 90%)
#> Parameter settings [===============================================================================================] 10/10 (100%)
#> CV fold: Fold3
#> CV progress [========================================================================================================] 3/3 (100%)
#>
#> Parameter settings [==================>-----------------------------------------------------------------------------] 2/10 ( 20%)
#> Parameter settings [============================>-------------------------------------------------------------------] 3/10 ( 30%)
#> Parameter settings [=====================================>----------------------------------------------------------] 4/10 ( 40%)
#> Parameter settings [===============================================>------------------------------------------------] 5/10 ( 50%)
#> Parameter settings [=========================================================>--------------------------------------] 6/10 ( 60%)
#> Parameter settings [==================================================================>-----------------------------] 7/10 ( 70%)
#> Parameter settings [============================================================================>-------------------] 8/10 ( 80%)
#> Parameter settings [=====================================================================================>----------] 9/10 ( 90%)
#> Parameter settings [===============================================================================================] 10/10 (100%)
head(validator_results)
#> fold performance lambda alpha family type.measure standardize
#> 1: Fold1 0.3465038 0.006548214 0.90 multinomial class TRUE
#> 2: Fold2 0.3475436 0.001710793 0.65 multinomial class TRUE
#> 3: Fold3 0.3514970 0.038236018 0.10 multinomial class TRUE
<- mlexperiments::MLNestedCV$new(
validator learner = mllrnrs::LearnerGlmnet$new(
metric_optimization_higher_better = FALSE
),strategy = "bayesian",
fold_list = fold_list,
k_tuning = 3L,
ncores = ncores,
seed = 312
)
$parameter_grid <- parameter_grid
validator$learner_args <- learner_args
validator$split_type <- "stratified"
validator
$parameter_bounds <- parameter_bounds
validator$optim_args <- optim_args
validator
$predict_args <- predict_args
validator$performance_metric <- performance_metric
validator$performance_metric_args <- performance_metric_args
validator$return_models <- TRUE
validator
$set_data(
validatorx = train_x,
y = train_y
)
<- validator$execute()
validator_results #>
#> CV fold: Fold1
#>
#> Registering parallel backend using 4 cores.
#>
#> CV fold: Fold2
#> CV progress [====================================================================>-----------------------------------] 2/3 ( 67%)
#>
#> Registering parallel backend using 4 cores.
#>
#> CV fold: Fold3
#> CV progress [========================================================================================================] 3/3 (100%)
#>
#> Registering parallel backend using 4 cores.
head(validator_results)
#> fold performance alpha lambda family type.measure standardize
#> 1: Fold1 0.3652609 0.7000000 0.004817737 multinomial class TRUE
#> 2: Fold2 0.3416288 0.4178147 0.017108341 multinomial class TRUE
#> 3: Fold3 0.3467740 0.1000000 0.041963982 multinomial class TRUE
<- mlexperiments::predictions(
preds_glmnet object = validator,
newdata = test_x
)
<- mlexperiments::performance(
perf_glmnet object = validator,
prediction_results = preds_glmnet,
y_ground_truth = test_y
)
perf_glmnet#> model performance
#> 1: Fold1 0.3606304
#> 2: Fold2 0.3603913
#> 3: Fold3 0.3529292
Here, glmnet
’s weights
-argument is used to rescale the case-weights during the training.
# define the target weights
<- ifelse(train_y == "n", 0.8, ifelse(train_y == "ei", 1.2, 1))
y_weights head(y_weights)
#> [1] 1.2 1.2 0.0 0.8 0.8 0.0
<- mlexperiments::MLTuneParameters$new(
tuner_w_weights learner = mllrnrs::LearnerGlmnet$new(
metric_optimization_higher_better = FALSE
),strategy = "grid",
ncores = ncores,
seed = seed
)
$parameter_grid <- parameter_grid
tuner_w_weights$learner_args <- c(
tuner_w_weights
learner_args,list(case_weights = y_weights)
)$split_type <- "stratified"
tuner_w_weights
$set_data(
tuner_w_weightsx = train_x,
y = train_y
)
<- tuner_w_weights$execute(k = 3)
tuner_results_grid #>
#> Parameter settings [============================>-------------------------------------------------------------------] 3/10 ( 30%)
#> Parameter settings [=====================================>----------------------------------------------------------] 4/10 ( 40%)
#> Parameter settings [===============================================>------------------------------------------------] 5/10 ( 50%)
#> Parameter settings [=========================================================>--------------------------------------] 6/10 ( 60%)
#> Parameter settings [==================================================================>-----------------------------] 7/10 ( 70%)
#> Parameter settings [============================================================================>-------------------] 8/10 ( 80%)
#> Parameter settings [=====================================================================================>----------] 9/10 ( 90%)
#> Parameter settings [===============================================================================================] 10/10 (100%)
head(tuner_results_grid)
#> setting_id metric_optim_mean lambda alpha family type.measure standardize
#> <int> <num> <num> <num> <char> <char> <lgcl>
#> 1: 1 0.5428029 0.015786209 0.70 multinomial class TRUE
#> 2: 2 0.5410926 0.005314924 0.90 multinomial class TRUE
#> 3: 3 0.5425178 0.017000533 0.65 multinomial class TRUE
#> 4: 4 0.5429929 0.027372552 0.10 multinomial class TRUE
#> 5: 5 0.5422328 0.020387093 0.45 multinomial class TRUE
#> 6: 6 0.5428979 0.034381521 0.05 multinomial class TRUE
<- mlexperiments::MLCrossValidation$new(
validator learner = mllrnrs::LearnerGlmnet$new(
metric_optimization_higher_better = FALSE
),fold_list = fold_list,
ncores = ncores,
seed = seed
)
# append the optimized setting from above with the newly created weights
$learner_args <- c(
validator$results$best.setting[-1],
tunerlist("case_weights" = y_weights)
)
$predict_args <- predict_args
validator$performance_metric <- performance_metric
validator$performance_metric_args <- performance_metric_args
validator$return_models <- return_models
validator
$set_data(
validatorx = train_x,
y = train_y
)
<- validator$execute()
validator_results #>
#> CV fold: Fold1
#>
#> CV fold: Fold2
#>
#> CV fold: Fold3
#> CV progress [========================================================================================================] 3/3 (100%)
#>
head(validator_results)
#> fold performance alpha lambda family type.measure standardize
#> <char> <num> <num> <num> <char> <char> <lgcl>
#> 1: Fold1 0.4139240 0.7 0.003092562 multinomial class TRUE
#> 2: Fold2 0.3570648 0.7 0.003092562 multinomial class TRUE
#> 3: Fold3 0.3831881 0.7 0.003092562 multinomial class TRUE