Bootstrap aggregating with quasi-randomized layer (regression)

Parameters description can be found at https://techtonique.github.io/nnetsauce/

RandomBagRegressor(
  obj,
  n_estimators = 10L,
  n_hidden_features = 1L,
  activation_name = "relu",
  a = 0.01,
  nodes_sim = "sobol",
  bias = TRUE,
  dropout = 0,
  direct_link = FALSE,
  n_clusters = 2L,
  cluster_encode = TRUE,
  type_clust = "kmeans",
  col_sample = 1,
  row_sample = 1,
  n_jobs = NULL,
  seed = 123L,
  verbose = 1L,
  backend = c("cpu", "gpu", "tpu")
)

Examples

library(datasets)

n <- 20 ; p <- 5
X <- matrix(rnorm(n * p), n, p) # no intercept!
y <- rnorm(n)

obj <- sklearn$tree$DecisionTreeRegressor()
obj2 <- RandomBagRegressor(obj)
obj2$fit(X[1:12,], y[1:12])
#> RandomBagRegressor(col_sample=1.0, dropout=0.0, obj=DecisionTreeRegressor(),
#>                    row_sample=1.0)
print(obj2$score(X[13:20, ], y[13:20]))
#> [1] -1.770562