Model stack and data stack print methods no longer raise conditions to print to the console (#228).
Added missing commas and addressed formatting issues throughout the vignettes and articles. Backticks for package names were removed and missing parentheses for functions were added (@Joscelinrocha, #218).
Addressed inflation of butchered model stack object size after saving and reloading (#214).
Fixed type-checking bug for add_candidates(name).
Introduced support for parallel processing using the future framework. The stacks package previously supported parallelism with foreach, and users can use either framework for now. In a future release, stacks will begin the deprecation cycle for parallelism with foreach, so we encourage users to begin migrating their code now. See the Parallel Processing section in the tune package's "Optimizations" article to learn more (#866).
Improved error message for unsupported model modes (#152).
Refine package alt text (#202).
Update example objects, resolving deprecation warnings from recipes (#203).
Fix bug in type checking for blend_predictions(mixture) (#204).
Resolve package-level documentation aliasing notice from CRAN.
Added an augment() method for model_stack objects (#173).
Converted all character variables in the tree_frogs example data to factor
and updated downstream example objects (#177).
Fixed bug that resulted in errors when using model formulas with the
"mgcv" engine (#193).
Made several optimizations to reduce evaluation time and memory allocation when stacking.
Various bug fixes and improvements to documentation.
Removes an unneeded data import attribute from the tree_frogs example data
and its associated objects (#148).
blend_predictions() doesn't error anymore if the control argument isn't a
control_grid object. As long as the object passed to control
include the same elements as control_grid() output,
parsnip::condense_control() will handle input (#149).
Tightened integration with the workflowsets package (#161, #165).
Revamped errors, warnings, and messages. Prompts now provide more thorough context about where they arose, include more extensive references to documentation, and are correctly pluralized (#150, #167).
Various bug fixes and improvements to documentation.
stacks 1.0.0 is the first production release of the package. While this release includes only a few minor bug fixes, it's accompanied by a white paper recently published in the Journal of Open Source software. You can read that paper here!
This release:
type argument to predict is set to "class",
and the outcome levels differ from alphabetical order.This is a GitHub-only release and does not change package source code. This
update includes a data-raw/paper subdirectory containing source for a
contributed paper to the Journal of Open Source Software.
add_candidates (#99).times argument to blend_predictions that is passed on to
rsample::bootstraps when fitting stacking coefficients. Reducing this
argument from its default (25) greatly reduces the run time of
blend_predictions (#94).fit_members(), if available, and fail informatively if not (#118)."class" (#125).collect_parameters failing to return stacking coefficients
in the two-class classification setting.add_variables()
preprocessor.This release of the package changes some elements of the internal structure of model stacks. As such, model stacks stored as saved objects will need to be regenerated before predicting, plotting, printing, etc.
mixture argument to blend_predictions.workflow_map objects
from the new {workflowsets} package. The interface to add_candidates
for doing so is the same as with tune_results objects, and
add_candidates is now a generic function.make.names for associated candidate
members.fit_members() will now warn when supplied a model stack whose
members have already been fitted.autoplot with type = "members"
more informative.Initial release!