Skip to contents

Bagging over LCG-seeded bootstrap replicates (Geron Ch 6, morie.fn hmbag)

Source: R/geron_ml_native.R
morie_geron_bagging.Rd

Bootstrap indices come from the exact-integer LCG, index m using seed

  • 7919 m, so replicates are byte-identical across machines. Default weak learner is a least-squares 1-split stump.

Usage

morie_geron_bagging(
  X,
  y,
  base_estimator = NULL,
  n_estimators = 10,
  seed = 0,
  task = "auto"
)

Arguments

X, y

Training data.

base_estimator

Optional function(Xb, yb).

n_estimators

Replicates.

seed

LCG seed.

task

"auto", "regression" or "classification".

Value

List with predict, train_pred, train_mse, oob_pred, oob_mse, oob_coverage, member_preds, task.

Examples

V <- c(1, 2, 3, 4, 5, 6, 7, 8)
morie_geron_bagging(V, V)
#> $predict
#> function (Xnew) 
#> {
#>     Bm <- if (is.matrix(Xnew)) 
#>         Xnew
#>     else matrix(as.numeric(Xnew), ncol = ncol(A))
#>     .morie_gr_need(ncol(Bm) == ncol(A), "predict: feature count mismatch")
#>     P <- do.call(rbind, lapply(models, function(f) as.numeric(f(Bm))))
#>     aggregate(P)
#> }
#> <environment: 0x56216a5497c0>
#> 
#> $train_pred
#> [1] 2.495000 2.495000 2.495000 3.353095 5.631429 6.271429 6.271429 6.271429
#> 
#> $train_mse
#> [1] 0.893089
#> 
#> $oob_pred
#> [1] 3.600000 2.550000 2.000000 3.241071 5.526984       NA 5.766667 6.082993
#> 
#> $oob_mse
#> [1] 2.01603
#> 
#> $oob_coverage
#> [1] 0.875
#> 
#> $estimators
#> $estimators[[1]]
#> function (Anew) 
#> {
#>     Bm <- if (is.matrix(Anew)) 
#>         Anew
#>     else matrix(as.numeric(Anew), ncol = ncol(A))
#>     ifelse(Bm[, j] <= thr, lp, rp)
#> }
#> <environment: 0x56216a514860>
#> 
#> $estimators[[2]]
#> function (Anew) 
#> {
#>     Bm <- if (is.matrix(Anew)) 
#>         Anew
#>     else matrix(as.numeric(Anew), ncol = ncol(A))
#>     ifelse(Bm[, j] <= thr, lp, rp)
#> }
#> <bytecode: 0x56215cd56b80>
#> <environment: 0x562151ee2ce8>
#> 
#> $estimators[[3]]
#> function (Anew) 
#> {
#>     Bm <- if (is.matrix(Anew)) 
#>         Anew
#>     else matrix(as.numeric(Anew), ncol = ncol(A))
#>     ifelse(Bm[, j] <= thr, lp, rp)
#> }
#> <bytecode: 0x56215cd56b80>
#> <environment: 0x56215b21cb00>
#> 
#> $estimators[[4]]
#> function (Anew) 
#> {
#>     Bm <- if (is.matrix(Anew)) 
#>         Anew
#>     else matrix(as.numeric(Anew), ncol = ncol(A))
#>     ifelse(Bm[, j] <= thr, lp, rp)
#> }
#> <bytecode: 0x56215cd56b80>
#> <environment: 0x56215b77e728>
#> 
#> $estimators[[5]]
#> function (Anew) 
#> {
#>     Bm <- if (is.matrix(Anew)) 
#>         Anew
#>     else matrix(as.numeric(Anew), ncol = ncol(A))
#>     ifelse(Bm[, j] <= thr, lp, rp)
#> }
#> <bytecode: 0x56215cd56b80>
#> <environment: 0x5621536c75e8>
#> 
#> $estimators[[6]]
#> function (Anew) 
#> {
#>     Bm <- if (is.matrix(Anew)) 
#>         Anew
#>     else matrix(as.numeric(Anew), ncol = ncol(A))
#>     ifelse(Bm[, j] <= thr, lp, rp)
#> }
#> <bytecode: 0x56215cd56b80>
#> <environment: 0x5621566e4d40>
#> 
#> $estimators[[7]]
#> function (Anew) 
#> {
#>     Bm <- if (is.matrix(Anew)) 
#>         Anew
#>     else matrix(as.numeric(Anew), ncol = ncol(A))
#>     ifelse(Bm[, j] <= thr, lp, rp)
#> }
#> <bytecode: 0x56215cd56b80>
#> <environment: 0x562161deeab8>
#> 
#> $estimators[[8]]
#> function (Anew) 
#> {
#>     Bm <- if (is.matrix(Anew)) 
#>         Anew
#>     else matrix(as.numeric(Anew), ncol = ncol(A))
#>     ifelse(Bm[, j] <= thr, lp, rp)
#> }
#> <bytecode: 0x56215cd56b80>
#> <environment: 0x562164593ed0>
#> 
#> $estimators[[9]]
#> function (Anew) 
#> {
#>     Bm <- if (is.matrix(Anew)) 
#>         Anew
#>     else matrix(as.numeric(Anew), ncol = ncol(A))
#>     ifelse(Bm[, j] <= thr, lp, rp)
#> }
#> <bytecode: 0x56215cd56b80>
#> <environment: 0x5621587b2e28>
#> 
#> $estimators[[10]]
#> function (Anew) 
#> {
#>     Bm <- if (is.matrix(Anew)) 
#>         Anew
#>     else matrix(as.numeric(Anew), ncol = ncol(A))
#>     ifelse(Bm[, j] <= thr, lp, rp)
#> }
#> <bytecode: 0x56215cd56b80>
#> <environment: 0x562153837650>
#> 
#> 
#> $member_preds
#>           [,1]     [,2]     [,3]     [,4]     [,5]     [,6]     [,7]     [,8]
#>  [1,] 3.000000 3.000000 3.000000 3.000000 6.000000 6.000000 6.000000 6.000000
#>  [2,] 2.000000 2.000000 2.000000 2.000000 6.000000 6.000000 6.000000 6.000000
#>  [3,] 2.000000 2.000000 2.000000 2.000000 7.200000 7.200000 7.200000 7.200000
#>  [4,] 3.000000 3.000000 3.000000 3.000000 6.333333 6.333333 6.333333 6.333333
#>  [5,] 2.500000 2.500000 2.500000 2.500000 6.000000 6.000000 6.000000 6.000000
#>  [6,] 2.250000 2.250000 2.250000 2.250000 6.000000 6.000000 6.000000 6.000000
#>  [7,] 2.666667 2.666667 2.666667 2.666667 2.666667 6.600000 6.600000 6.600000
#>  [8,] 2.333333 2.333333 2.333333 5.200000 5.200000 5.200000 5.200000 5.200000
#>  [9,] 4.200000 4.200000 4.200000 4.200000 4.200000 6.666667 6.666667 6.666667
#> [10,] 1.000000 1.000000 1.000000 6.714286 6.714286 6.714286 6.714286 6.714286
#> 
#> $task
#> [1] "regression"
#> 
#> $estimate
#> [1] 0.893089
#> 
#> $n
#> [1] 8
#> 
#> $method
#> [1] "Bagging over LCG-seeded bootstrap replicates"
#>