R package: bug #41
Description
When I run this code, I expect that when max_depth = inf, min_sample = 1, and margin = 1, with all interval lengths being 2, the prediction for each training point should be the midpoint of the target interval. However, this code does not produce what I expected.
n_points <- 10
feature.mat <- data.frame(x = 1:n_points)
min_target <- rep(c(0, 3), length.out = n_points)
max_target <- rep(c(2, 5), length.out = n_points)
target.mat <- cbind(min_target, max_target)
tree <- mmit(feature.mat = feature.mat, target.mat = target.mat, max_depth = Inf, min_sample = 1, margin = 1, loss = 'hinge')`