finetune 1.1.0
CRAN release: 2023-04-19
Various minor changes to keep up with developments in the tune and dplyr packages (#60) (#62) (#67) (#68).
Corrects
.config
output withsave_pred = TRUE
intune_sim_anneal()
. The function previously outputted a constantModel1_Preprocessor1
in the.predictions
slot, and now provides.config
values that align with those in.metrics
(#57).
finetune 1.0.1
CRAN release: 2022-10-12
For racing:
collect_metrics()
andcollect_predictions()
have a'complete'
argument that only returns results for model configurations that were fully resampled.select_best()
andshow_best()
now only show results for model configurations that were fully resampled.tune_race_anova()
,tune_race_win_loss()
, andtune_sim_anneal()
no longer error ifcontrol
argument isn’t a the correspondingcontrol_*()
object. Will work as long as the object passed tocontrol
includes the same elements as the requiredcontrol_*()
object.The
control_sim_anneal()
got a new argumentverbose_iter
that is used to control the verbosity of the iterative calculations. This change means that theverbose
argument is being passed totune_grid()
to control its verbosity.
finetune 1.0.0
CRAN release: 2022-09-05
An informative error is given when there are not enough resamples for racing (#33).
tune_sim_anneal()
was not passing all arguments totune_grid()
(#40).
finetune 0.2.0
CRAN release: 2022-03-24
Maintenance release for CRAN requirements.
Use
extract_parameter_set_dials()
instead ofparameters()
to get parameter sets.Removed some pillar-related S3 methods that currently live in tune.
finetune 0.1.1
CRAN release: 2022-02-21
-
tune_sim_anneal()
only overwrites tuning parameter information when they originally contain unknowns.
finetune 0.1.0
CRAN release: 2021-07-21
A check was added to make sure that
lme4
orBradleyTerry2
are installed (#8)Added
pillar
methods for formattingtune
objects in list columns.Fixed bug in
random_integer_neighbor_calc()
to keep values inside range (#10)tune_sim_anneal()
now retains a finalized parameter set and replaces any existing parameter set that was not finalized (#14)A bug in win/loss racing was fixed for cases when one tuning parameter had results that were so bad that it broke the Bradley-Terry model (#7)