ML.XGBoost: ML.XGBoost
In frbl/OnlineSuperLearner: Online SuperLearner package

Base class for any XGBoost machine learning model.

1	ML.XGBoost

An object of class R6ClassGenerator of length 24.

initialize(booster = 'gblinear', max_depth = 6, nthread = 1, alpha = 0, lambda = 0, rounds = 200, gamma = 0, eta = 0.3, objective = 'binary:logistic', verbose = FALSE)

Initializes a new XGBoosted estimator. See the underlying xgboost packages for more details. This estimator allows to tweak several hyperparameters (see params). By default XGBoost uses elasticnet for penalizing the fitted model, the amount of penalization can be tweaked using the alpha (L1 regularization) and lambda (L2 regularization). See https://github.com/dmlc/xgboost/blob/master/doc/parameter.md

@param booster string (default = 'gblinear') the booster to use for fitting the booster. Can be either of gbtree, gblinear or dart.

@param max_depth integer (default = 6) the max depth of the GBM.

@param nthread integer (default = 1) the number of threads to run the XBoost algortihm on. Note!! Setting this to a different setting might cause unwanted behavior! If set to -1, it will use all cores available.

@param alpha double L1 regularization parameter

@param lambda double L2 regularization parameter

@param rounds = The number of rounds for boosting

@param gamma minimum loss reduction required to make a further partition on a leaf node of the tree. The larger, the more conservative the algorithm will be.

@param eta double (default = 0.3) the stepsize used

@param objective string (default = 'binary:logistic') the objective to optimize.

get_nthread

Active method. Function that returns the number of threads the XGBoost algorithm runs on.

get_validity

Active method. Function that shows wheter the current configuration of the booster is valid. The function returns TRUE if everything is specified correctly. It will throw an error (with the error messages) when something is misspecified. This function is automatically called after initialization.

frbl/OnlineSuperLearner documentation built on Feb. 9, 2020, 9:28 p.m.