Constructing a Learner
A learner is a machine learning algorithm implementation in the mlr package. As highlighted in the previous section on the mlr package, there is a rich collection of such learner functions in mlr.
For our scene classification problem, the mlr package offers building a multilabel classification model in two possible ways:
Adaptation method: In this, we adapt an explicit algorithm on the entire problem.
Transformation method: We transform the problem into a simple binary classification problem and then apply the available algorithm for the binary classification.
Adaptation Methods
The mlr package in R offers two algorithm adaption methods. First, the multivariate random forest algorithm that comes from the randomForestSRC package, and second, the random ferns multilabel algorithm built in the rFerns package.
The makeLearner() function in mlr creates the model object for the rFerns and randomForestSRC algorithms, as shown in the following code:
multilabel.lrn3 = makeLearner...
