f_m = f_(m-1) + nu T_m where each regression tree T_m is fitted to the
residuals, the negative gradient (ESL Alg. 10.3, eqs 10.29-10.41); starts
at mean(y). With stumps this equals gbm's gaussian boosting without
subsampling.
Usage
morie_esl_gbm(X, y, M = 100, nu = 0.1, max_depth = 2, min_leaf = 1)
Arguments
- X
Predictors.
- y
Response.
- M
Number of trees.
- nu
Shrinkage in (0, 1].
- max_depth, min_leaf
Tree controls.
Value
Named list: estimate (training RSS), f0, trees, nu, M,
train_rss_path, fitted, n, p.
References
Friedman, J. (2001). Annals of Statistics 29, 1189-1232.
Examples
i <- 1:60
X <- cbind(((7 * i) %% 59) / 59 * 3, ((11 * i) %% 61) / 61 * 2)
morie_esl_gbm(X, sin(2 * X[, 1]), M = 10, max_depth = 1)$estimate
#> [1] 7.43119
