Exercise

Perform hyperparameter tuning with mlr

Now, you can combine the prepared functions and objects from the previous exercise to actually perform hyperparameter tuning with random search. The knowledge_train_data dataset has already been loaded for you, as have the packages mlr, tidyverse and tictoc. And the following code has also been run already:

# Define task
task <- makeClassifTask(data = knowledge_train_data, 
                        target = "UNS")

# Define learner
lrn <- makeLearner("classif.nnet", predict.type = "prob", fix.factors.prediction = TRUE)

# Define set of parameters
param_set <- makeParamSet(
  makeDiscreteParam("size", values = c(2,3,5)),
  makeNumericParam("decay", lower = 0.0001, upper = 0.1)
)

Instructions 1/4

undefined XP
    1
    2
    3
    4
  • Change the maximum number of iterations for random search to 6. Note, that 6 is a very low number; we use it so that calculation won't take forever to complete here; usually, you would set the number much higher (the default is 100).