Infinite boosting rounds with "logloss" loss function

[Harrison4192]

I was trying to predict a binary variable specifying the loss_function as "logloss", but instead of stopping early, it seems the boosting rounds continue even while improvement stops. I expected training on the iris dataset to be very quick (like with loss_function "mse") but it just keeps going. The same issue is happening with other loss functions (similar issue with "poisson") except "mse" which performs as expected. The example can be reproduced as below.

[Blunde1]

@Harrison4192 Thanks!
I assume this is with agtboost 0.9.2?

[Blunde1]

Just confirmed it also happens with 0.9.1
Will look into this

library(fastDummies)
library(dplyr)
require(devtools)
install_version("agtboost", version = "0.9.1", repos = "http://cran.us.r-project.org")
library(agtboost)

iris %>%
  dummy_cols(remove_selected_columns=T) -> iris1

iris1 %>%
  select(-Species_setosa) %>%
  as.matrix -> iris2

mod <- gbt.train(
  y=iris1$Species_setosa,
  x=iris2,
  loss_function = "logloss",
  verbose=1,
  nrounds=300,
)

[Blunde1]

• Find issue
• Solve issue
• Setup tests in CI so that similar behavior will be detected in the future

[Harrison4192]

Hi, yea i can confirm this issue on 0.9.1 and the current development version 0.9.2