tests/testthat/test-varselect_hal.R

In hal9001: The Scalable Highly Adaptive Lasso

# context("HAL with screening for high-dimensional data")
# set.seed(45791)

# easily compute MSE
# mse <- function(preds, y) {
# mean((preds - y)^2)
# }

# generate simple test data
# n <- 1000
# p <- 100
# x <- xmat <- matrix(rnorm(n * p), n, p)
# y_prob <- plogis(3 * sin(x[, 1]) + 3 * sin(x[, 2]))
# y <- rbinom(n = n, size = 1, prob = y_prob)

# test_n <- 10000
# test_x <- matrix(rnorm(test_n * p), test_n, p)
# test_y_prob <- plogis(3 * sin(test_x[, 1]) + sin(test_x[, 2]))
# test_y <- rbinom(n = test_n, size = 1, prob = y_prob)

# col_lists <- as.list(1:p)
# i <- 1

# linear_glmnet <- suppressWarnings(glmnet(
# x = cbind(y, y), y = y, family = "binomial", maxit = 1,
# thresh = 0.01
# ))
# linear_glmnet$lambda

# TODO: test screening functionality