Survival DNNSurv Learner

Fits a neural network based on pseudo-conditional survival probabilities. Calls survivalmodels::dnnsurv() from package 'survivalmodels'.

Details

Custom nets can be used in this learner either using the survivalmodels::build_keras_net utility function or using keras. The number of output channels should be of length 1 and number of input channels is the number of features plus number of cuts.

Dictionary

This Learner can be instantiated via the dictionary mlr_learners or with the associated sugar function lrn():

mlr_learners$get("surv.dnnsurv")
lrn("surv.dnnsurv")

Meta Information

• Task type: “surv”

• Predict Types: “crank”, “distr”

• Feature Types: “integer”, “numeric”

• Required Packages: mlr3, mlr3proba, mlr3extralearners, survivalmodels, keras, pseudo, tensorflow, distr6

Parameters

Id	Type	Default	Levels	Range
cuts	integer	5		\([1, \infty)\)
cutpoints	untyped	-		-
custom_model	untyped	-		-
optimizer	character	adam	adadelta, adagrad, adamax, adam, nadam, rmsprop, sgd	-
lr	numeric	0.02		\([0, \infty)\)
beta_1	numeric	0.9		\([0, 1]\)
beta_2	numeric	0.999		\([0, 1]\)
epsilon	numeric	-		\([0, \infty)\)
decay	numeric	0		\([0, \infty)\)
clipnorm	numeric	-		\((-\infty, \infty)\)
clipvalue	numeric	-		\((-\infty, \infty)\)
schedule_decay	numeric	0.04		\((-\infty, \infty)\)
momentum	numeric	0		\([0, \infty)\)
nesterov	logical	FALSE	TRUE, FALSE	-
loss_weights	untyped	-		-
weighted_metrics	untyped	-		-
early_stopping	logical	FALSE	TRUE, FALSE	-
min_delta	numeric	0		\([0, \infty)\)
patience	integer	0		\([0, \infty)\)
verbose	integer	0		\([0, 2]\)
baseline	numeric	-		\((-\infty, \infty)\)
restore_best_weights	logical	FALSE	TRUE, FALSE	-
batch_size	integer	32		\([1, \infty)\)
epochs	integer	10		\([1, \infty)\)
validation_split	numeric	0		\([0, 1]\)
shuffle	logical	TRUE	TRUE, FALSE	-
sample_weight	untyped	-		-
initial_epoch	integer	0		\([0, \infty)\)
steps_per_epoch	integer	-		\([1, \infty)\)
validation_steps	integer	-		\([1, \infty)\)
steps	integer	-		\([0, \infty)\)
callbacks	untyped	-		-

Installation

Package 'survivalmodels' is not on CRAN and has to be install from GitHub via remotes::install_github("RaphaelS1/survivalmodels").

Custom mlr3 defaults

• verbose:

  • Actual default: 1L

  • Adjusted default: 0L

  • Reason for change: Prevents plotting.

References

Zhao, Lili, Feng, Dai (2019). “Dnnsurv: Deep neural networks for survival analysis using pseudo values.” arXiv preprint arXiv:1908.02337.

See also

• Dictionary of Learners: mlr3::mlr_learners.

• as.data.table(mlr_learners) for a table of available Learners in the running session (depending on the loaded packages).

• Chapter in the mlr3book: https://mlr3book.mlr-org.com/basics.html#learners

• mlr3learners for a selection of recommended learners.

• mlr3cluster for unsupervised clustering learners.

• mlr3pipelines to combine learners with pre- and postprocessing steps.

• mlr3tuning for tuning of hyperparameters, mlr3tuningspaces for established default tuning spaces.

Author

RaphaelS1

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvDNNSurv

Methods

Public methods

• LearnerSurvDNNSurv$new()

• LearnerSurvDNNSurv$clone()

Inherited methods

• mlr3::Learner$base_learner()
• mlr3::Learner$format()
• mlr3::Learner$help()
• mlr3::Learner$predict()
• mlr3::Learner$predict_newdata()
• mlr3::Learner$print()
• mlr3::Learner$reset()
• mlr3::Learner$train()

---

Method new()

Creates a new instance of this R6 class.

Usage

LearnerSurvDNNSurv$new()

---

Method clone()

The objects of this class are cloneable with this method.

Usage

LearnerSurvDNNSurv$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

learner = mlr3::lrn("surv.dnnsurv")
print(learner)
#> <LearnerSurvDNNSurv:surv.dnnsurv>: Neural Network
#> * Model: -
#> * Parameters: verbose=0
#> * Packages: mlr3, mlr3proba, mlr3extralearners, survivalmodels, keras,
#>   pseudo, tensorflow, distr6
#> * Predict Types:  [crank], distr
#> * Feature Types: integer, numeric
#> * Properties: -

# available parameters:
learner$param_set$ids()
#>  [1] "cuts"                 "cutpoints"            "custom_model"        
#>  [4] "optimizer"            "lr"                   "beta_1"              
#>  [7] "beta_2"               "epsilon"              "decay"               
#> [10] "clipnorm"             "clipvalue"            "schedule_decay"      
#> [13] "momentum"             "nesterov"             "loss_weights"        
#> [16] "weighted_metrics"     "early_stopping"       "min_delta"           
#> [19] "patience"             "verbose"              "baseline"            
#> [22] "restore_best_weights" "batch_size"           "epochs"              
#> [25] "validation_split"     "shuffle"              "sample_weight"       
#> [28] "initial_epoch"        "steps_per_epoch"      "validation_steps"    
#> [31] "steps"                "callbacks"