Understanding SMAC
SMAC is part of the BO hyperparameter tuning method group and utilizes random forest as the surrogate model. This method is optimized to handle discrete or categorical hyperparameters. If your hyperparameter space is huge and is dominated by discrete hyperparameters, then SMAC is a good choice for you.
Similar to BOGP, SMAC also works by modeling the objective function. Specifically, it utilizes random forest as the surrogate model to create an estimation of the real objective function, which can then be passed to the acquisition function (see the Introducing BO section for more details).
Random forest is a machine learning (ML) algorithm that can be utilized in classification or regression tasks. It is built upon a collection of decision trees, which is known to perform well with categorical types of features. The name random forest comes from the fact that it is built from several decision trees. We will discuss random forest, along with its hyperparameters...