# test_that("summary.bnns works for regression (linear output)", {
#   set.seed(123)
#   df <- data.frame(x1 = runif(10), x2 = runif(10), y = rnorm(10))
#
#   result <- summary.bnns(bnns(y ~ -1 + x1 + x2, data = df, L = 1, nodes = 2, act_fn = 5, out_act_fn = 1, iter = 1e1, warmup = 5, chains = 1))
#
#   expect_true(any(grepl("rmse", names(result$Performance))))
#   expect_type(result, "list")
#   expect_equal(result$`Number of observations`, nrow(df))
# })
#
# test_that("summary.bnns works for binary classification (sigmoid output)", {
#   set.seed(123)
#   df <- data.frame(x1 = runif(10), x2 = runif(10), y = sample(0:1, 10, replace = TRUE))
#
#   result <- summary.bnns(bnns(y ~ -1 + x1 + x2, data = df, L = 1, nodes = 2, act_fn = 5, out_act_fn = 2, iter = 1e1, warmup = 5, chains = 1))
#
#   expect_true(any(grepl("accuracy", names(result$Performance))))
#   expect_type(result, "list")
#   expect_equal(result$`Number of observations`, nrow(df))
# })
#
# test_that("summary.bnns works for multiclass classification (softmax output)", {
#   set.seed(123)
#   df <- data.frame(x1 = runif(10), x2 = runif(10), y = factor(sample(letters[1:3], 10, replace = TRUE)))
#
#   result <- summary.bnns(bnns(y ~ -1 + x1 + x2, data = df, L = 1, nodes = 2, act_fn = 5, out_act_fn = 3, iter = 1e1, warmup = 5, chains = 1))
#
#   expect_true(any(grepl("AUC", names(result$Performance))))
#   expect_type(result, "list")
#   expect_equal(result$`Number of observations`, nrow(df))
# })