use grid_space_filling() instead of grid_latin_hypercube() #919
Conversation
| @@ -632,7 +632,7 @@ fit_gp <- function(dat, pset, metric, eval_time = NULL, control, ...) { | |||
|
|
|||
| pred_gp <- function(object, pset, size = 5000, current = NULL, control) { | |||
| pred_grid <- | |||
| dials::grid_latin_hypercube(pset, size = size) %>% | |||
| dials::grid_space_filling(pset, size = size, type = "latin_hypercube") %>% | |||
There was a problem hiding this comment.
I stuck with a LHD here because the new designs take a really long time to compute a really large grid (5,000 points).
| 7 <split [29/3]> Fold07 <tibble [6 x 5]> <tibble [0 x 4]> | ||
| 8 <split [29/3]> Fold08 <tibble [6 x 5]> <tibble [0 x 4]> | ||
| 9 <split [29/3]> Fold09 <tibble [6 x 5]> <tibble [0 x 4]> | ||
| 10 <split [29/3]> Fold10 <tibble [6 x 5]> <tibble [0 x 4]> |
There was a problem hiding this comment.
I'm surprised that this snap is changing. What's the explanation here?
There was a problem hiding this comment.
The "6" is because we have three standard metrics (added brier) but also different due to different initial grids.
There was a problem hiding this comment.
I understand that we have 3 default classification metrics now, but why wouldn't that snapped have failed before this PR? What do the grids have to do with this change?
|
This pull request has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex: https://reprex.tidyverse.org) and link to this issue. |
Depends on dev dials