XGBoost regularized objective (R parity) — morie_xgboost

Wraps the xgboost package. If xgboost isn't installed, falls back to gbm (gradient boosting) so users still get a usable boosted-trees result; the backend is flagged in the output.

Usage

morie_xgboost_objective(
  x,
  y,
  n_estimators = 100L,
  learning_rate = 0.1,
  max_depth = 3L,
  reg_lambda = 1,
  reg_alpha = 0,
  task = "auto",
  seed = 0L,
  deterministic_seed = NULL
)

Arguments

x: Numeric predictor matrix.
y: Response.
n_estimators: Number of boosting rounds.
learning_rate: eta / shrinkage.
max_depth: Tree depth.
reg_lambda: L2 leaf penalty.
reg_alpha: L1 leaf penalty.
task: "auto", "classification", or "regression".
seed: RNG seed.
deterministic_seed: Integer or NULL. If supplied, the RNG state is derived from the SHA-keyed morie_det_rng() so Py<->R streams agree on the canonical fixture. When NULL (default), behaviour is unchanged: seed drives set.seed() directly.

Value

Named list: estimate, train_score, feature_importances, backend, n_estimators, learning_rate, max_depth, reg_lambda, reg_alpha, task, n, method.

Examples

morie_xgboost_objective(x = rnorm(50), y = rnorm(50))
#> $estimate
#> [1] 0.8654083
#> 
#> $train_score
#> [1] 0.8654083
#> 
#> $feature_importances
#> [1] 0.4503273 0.5496727
#> 
#> $backend
#> [1] "xgboost"
#> 
#> $n_estimators
#> [1] 100
#> 
#> $learning_rate
#> [1] 0.1
#> 
#> $max_depth
#> [1] 3
#> 
#> $reg_lambda
#> [1] 1
#> 
#> $reg_alpha
#> [1] 0
#> 
#> $task
#> [1] "regression"
#> 
#> $n
#> [1] 50
#> 
#> $method
#> [1] "XGBoost-style boosting (xgboost, regression)"
#>