Encapsulate a Graph as a Learner

A Learner that encapsulates a Graph to be used in mlr3 resampling and benchmarks.

The Graph must return a single Prediction on its $predict() call. The result of the $train() call is discarded, only the internal state changes during training are used.

The predict_type of a GraphLearner can be obtained or set via it's predict_type active binding. Setting a new predict type will try to set the predict_type in all relevant PipeOp / Learner encapsulated within the Graph. Similarly, the predict_type of a Graph will always be the smallest denominator in the Graph.

A GraphLearner is always constructed in an untrained state. When the graph argument has a non-NULL $state, it is ignored.

Format

R6Class object inheriting from mlr3::Learner.

Construction

GraphLearner$new(graph, id = NULL, param_vals = list(), task_type = NULL, predict_type = NULL)

graph :: Graph | PipeOp
Graph to wrap. Can be a PipeOp, which is automatically converted to a Graph. This argument is usually cloned, unless clone_graph is FALSE; to access the Graph inside GraphLearner by-reference, use $graph.
id :: character(1) Identifier of the resulting Learner.
param_vals :: named list
List of hyperparameter settings, overwriting the hyperparameter settings . Default list().
task_type :: character(1)
What task_type the GraphLearner should have; usually automatically inferred for Graphs that are simple enough.
predict_type :: character(1)
What predict_type the GraphLearner should have; usually automatically inferred for Graphs that are simple enough.
clone_graph :: logical(1)
Whether to clone graph upon construction. Unintentionally changing graph by reference can lead to unexpected behaviour, so TRUE (default) is recommended. In particular, note that the $state of $graph is set to NULL by reference on construction of GraphLearner, during $train(), and during $predict() when clone_graph is FALSE.

Fields

Fields inherited from PipeOp, as well as:

graph :: Graph
Graph that is being wrapped. This field contains the prototype of the Graph that is being trained, but does not contain the model. Use graph_model to access the trained Graph after $train(). Read-only.
graph_model :: Learner
Graph that is being wrapped. This Graph contains a trained state after $train(). Read-only.

Internals

as_graph() is called on the graph argument, so it can technically also be a list of things, which is automatically converted to a Graph via gunion(); however, this will usually not result in a valid Graph that can work as a Learner. graph can furthermore be a Learner, which is then automatically wrapped in a Graph, which is then again wrapped in a GraphLearner object; this usually only adds overhead and is not recommended.

Examples

library("mlr3")

graph = po("pca") %>>% lrn("classif.rpart")

lr = GraphLearner$new(graph)
lr = as_learner(graph)  # equivalent

lr$train(tsk("iris"))

lr$graph$state  # untrained version!
#> $pca
#> NULL
#> 
#> $classif.rpart
#> NULL
#> 
# The following is therefore NULL:
lr$graph$pipeops$classif.rpart$learner_model$model
#> NULL

# To access the trained model from the PipeOpLearner's Learner, use:
lr$graph_model$pipeops$classif.rpart$learner_model$model
#> n= 150 
#> 
#> node), split, n, loss, yval, (yprob)
#>       * denotes terminal node
#> 
#> 1) root 150 100 setosa (0.33333333 0.33333333 0.33333333)  
#>   2) PC1< -1.553145 50   0 setosa (1.00000000 0.00000000 0.00000000) *
#>   3) PC1>=-1.553145 100  50 versicolor (0.00000000 0.50000000 0.50000000)  
#>     6) PC1< 1.142805 44   1 versicolor (0.00000000 0.97727273 0.02272727) *
#>     7) PC1>=1.142805 56   7 virginica (0.00000000 0.12500000 0.87500000) *

# Feature importance (of principal components):
lr$graph_model$pipeops$classif.rpart$learner_model$importance()
#>       PC1       PC2       PC3       PC4 
#> 85.795455 18.016529  5.694731  3.254132

Format

Construction

Fields

Internals

See also

Examples