Randomized search with K-fold CV (Geron Ch 2, morie.fn grrnd)

The LCG stream is consumed first for the fold permutation (m draws, argsorted stably) and then one draw PER PARAMETER per configuration, in the order of param_dist. std_score is the ddof = 0 deviation.

Usage

morie_geron_randomized_search_cv(
  X,
  y,
  param_dist,
  n_iter,
  K,
  fit_score = NULL,
  seed = 42
)

Arguments

X, y

Data.

param_dist

Named list of specs: a function(u), a length-2 numeric range, or a vector of discrete choices.

n_iter

Configurations to draw.

K

Folds in [2, m].

fit_score

Required scoring function.

seed

LCG seed.

Value

List with best_params, best_score, results, fold_sizes.

Examples

if (requireNamespace("ranger", quietly = TRUE)) {
  set.seed(3)
  X <- matrix(rnorm(60), 30, 2)
  y <- as.numeric(X[, 1] > 0)
  fs <- function(Xtr, ytr, Xte, yte, params) {
    b <- qr.solve(cbind(1, as.matrix(Xtr)), as.numeric(ytr))
    pred <- as.numeric(cbind(1, as.matrix(Xte)) %*% b)
    -mean((pred - yte)^2) * params$w
  }
  morie_geron_randomized_search_cv(X, y, param_dist = list(w = c(0.5, 2)),
                                   n_iter = 3, K = 3, fit_score = fs)
}
#> $best_params
#> $best_params$w
#> [1] 0.5336829
#> 
#> 
#> $best_score
#> [1] -0.04705493
#> 
#> $results
#> $results[[1]]
#> $results[[1]]$params
#> $results[[1]]$params$w
#> [1] 0.5336829
#> 
#> 
#> $results[[1]]$mean_score
#> [1] -0.04705493
#> 
#> $results[[1]]$std_score
#> [1] 0.01707143
#> 
#> $results[[1]]$fold_scores
#> [1] -0.04621495 -0.06837041 -0.02657944
#> 
#> 
#> $results[[2]]
#> $results[[2]]$params
#> $results[[2]]$params$w
#> [1] 1.305933
#> 
#> 
#> $results[[2]]$mean_score
#> [1] -0.1151444
#> 
#> $results[[2]]$std_score
#> [1] 0.04177414
#> 
#> $results[[2]]$fold_scores
#> [1] -0.11308896 -0.16730383 -0.06504045
#> 
#> 
#> $results[[3]]
#> $results[[3]]$params
#> $results[[3]]$params$w
#> [1] 1.74494
#> 
#> 
#> $results[[3]]$mean_score
#> [1] -0.1538518
#> 
#> $results[[3]]$std_score
#> [1] 0.05581709
#> 
#> $results[[3]]$fold_scores
#> [1] -0.15110535 -0.22354529 -0.08690468
#> 
#> 
#> 
#> $fold_sizes
#> [1] 10 10 10
#> 
#> $estimate
#> [1] -0.04705493
#> 
#> $n
#> [1] 30
#>