Explanation of hyperparameter tuning procedure for regression tree ensembles

What regression tree ensemble methods and what parameters does Matlab actually consider in hyperparameter tuning?

See https://se.mathworks.com/help/stats/fitrensemble.html and the example "Optimize Regression Ensemble" therein. It says "You can find hyperparameters that minimize five-fold cross-validation loss by using automatic hyperparameter optimization."

But what is the search space here?

The output in that example only displays Bag and LSBoost as methods. Does it neglect random forests, i.e. subset sampling instead of bootstrapping the input space? Or is bagging here an umbrella term that covers also Random Forests?

Furthermore, the output in the above example only displays NumLearnCycles (tree count), LearnRate (for boosting) and MinLeafSize (obvious). How about treatment of the other CART decision tree algorithm hyperparameters? Are they included as default values - if so, then where to find them?

Answer 1

[Alan Weiss]

You can find all the information later on in that same reference page:

Hyperparameter Optimization Options

Alan Weiss

MATLAB mathematical toolbox documentation