gbt.convergence: Convergence of agtboost model.

In agtboost: Adaptive and Automatic Gradient Boosting Computations

Description

gbt.convergence calculates loss of data over iterations in the model

Usage

gbt.convergence(object, y, x)

Arguments

object	Object or pointer to object of class ENSEMBLE
y	response vector
x	design matrix for training. Must be of type matrix.

Details

Computes the loss on supplied data at each boosting iterations of the model passed as object. This may be used to visually test for overfitting on test data, or the converce, to check for underfitting or non-convergence.

Value

vector with $K+1$ elements with loss at each boosting iteration and at the first constant prediction

Examples

## Gaussian regression:
x_tr <- as.matrix(runif(500, 0, 4))
y_tr <- rnorm(500, x_tr, 1)
x_te <- as.matrix(runif(500, 0, 4))
y_te <- rnorm(500, x_te, 1)
mod <- gbt.train(y_tr, x_tr)
convergence <- gbt.convergence(mod, y_te, x_te)
which.min(convergence) # Should be fairly similar to boosting iterations + 1
mod$get_num_trees() +1 # num_trees does not include initial prediction

agtboost documentation built on Nov. 24, 2021, 1:16 a.m.