dials 1.4.3

New parameters

• ordinal_link() and odds_link() are two new parameters for the new ordinal_reg() models in parsnip (@corybrunson, #435).

• num_estimators(), softmax_temperature(), balance_probabilities(), average_before_softmax(), and training_set_limit() are new parameters for the tab_pfn() model in parsnip (#412).

Other changes

• grid_space_filling() now also respects the original argument when pre-made designs are available (#409).

• parameters() and the grid_*() functions give more information in the error message when non-parameter objects are passed in (#437, #438).

• encode_unit() now provides a helpful error when x is not a parameter object (#430).

• value_validate(), value_transform(), value_inverse(), and value_set() now produce more informative error messages when values contain unknowns (#445).

• value_validate(), encode_unit(), get_p(), get_log_p(), get_n(), get_n_frac(), get_n_frac_range(), and get_batch_sizes() now enforce empty dots (#439).

• Several further type-checking error messages now include the actual type of the input (#423).

dials 1.4.2

• prop_terms() is a new parameter object used for recipes that do supervised feature selection (#395).

• upper_limit() and lower_limit() now have ranges that are inclusive of the endpoints, unless the endpoint is infinite (#396).

• batch_size() now has a specific default parameter range instead of an unknown default range. get_batch_sizes() is deprecated (#398).

dials 1.4.1

• Two new parameters, cal_method_class() and cal_method_reg(), to control which method is used to calibrate model predictions (#383).

• regularization_factor() is now exclusive of the lower border 0 in compliance with ranger::ranger() (#381).

dials 1.4.0

• For space-filling designs for \(p\) parameters, there is a higher likelihood of finding a space-filling design for 1 < size <= p. Also, single-point designs now default to a random grid (#363).

• value_seq() and value_sample() now respect the inclusive argument of quantitative parameters (#347).

• The constructors, new_*_parameter(), now label unlabeled parameter (i.e., constructed with label = NULL) as such (#349).

• All messages, warnings and errors has been translated to use {cli} package (#311).

• parameters.list() now enforces the unused ellipsis to be empty (#378).

• Added three new parameters for use in postprocessing in the tailor package (#357).

  • buffer() sets the distance on either side of a classification threshold within which predictions are considered equivocal in tailor::adjust_equivocal_zone().
  • lower_limit() and upper_limit() sets the ranges for numeric predictions in tailor::adjust_numeric_range().

Breaking changes

• The grid_*() functions now error instead of warn when provided with the wrong argument to control the grid size. So grid_space_filling(), grid_random(), grid_max_entropy(), and grid_latin_hypercube() now error if used with a levels argument and grid_regular() now errors if used with a size argument (#368).

• The "optimal" option for the weight_func() parameter has been removed since it is choosing the optimal value based on the resubstition error (#370).

• When constructing integer-valued parameters with a range of two consecutive values the inclusive argument needs to be set to c(TRUE, TRUE) to leave at least two values to sample from (#373).

dials 1.3.0

Improvements

• The space-filling design functionality was expanded to include several new types of designs: Audze-Eglais, max/min L1, max/min L2, and uniform. These are all pre-computed designs accessed from the sfd package (#329).

• A new function is used to access all of the space-filling designs called grid_space_filling() (#329).

• Two new parameters, activation_2() and hidden_units_2(), for use with brulee::brulee_mlp_two_layer() (#339).

Deprecations

• grid_max_entropy() and grid_latin_hypercube() are deprecated in favor of grid_space_filling() (#332).

• pull_dials_object() has been removed (#344).

• The grid_*() methods for workflow objects have been removed (#344).

• The deprecation of the default argument to the constructors new_quant_param() and new_qual_param() has been escalated to an error (#344).

dials 1.2.1

New parameters

• Added initial_umap() and target_weight() for recipes::step_umap() (#324).

Other changes

• Improving styling of error messages by switching to cli (#315, #317, #321).

• Update usage of tranformation functions from the scales package to their new names and require the corresponding version v1.3.0 (#323).

dials 1.2.0

New parameters

• Added trim_amount() for recipes::step_impute_mean().

• Added num_runs() for recipes::step_nnmf() (#281).

• Added harmonic_frequency() for recipes::step_harmonic() (#281).

• Added validation_set_prop() for embed::step_discretize_xgb() (#280).

Other changes

• Deprecation of pull_dials_object() has been escalated to an error. Please use extract_parameter_dials() instead (#265).

• The methods grid_regular.workflow(), grid_random.workflow(), grid_max_entropy.workflow(), and grid_latin_hypercube.workflow() have been deprecated (#302).

• The constructor functions for single parameters, new_quant_param() and new_qual_param(), as well as for parameter sets, parameters_constr(), now have improved handling of the call shown in error messages (#291, #295).

• The constructor for parameter sets, parameters_constr(), now checks that all inputs have the same length (#295).

dials 1.1.0

New parameters

• Added learning rate scheduler parameters rate_decay(), rate_initial(), rate_largest(), rate_reduction(), rate_schedule(), rate_step_size(), and rate_steps() for the new brulee functions (#253).

• Added num_clusters() parameter for tidyclust models (#259).

• Added num_leaves() parameter for lightbgm models (@joeycouse, #256).

Other changes

• The default argument to the constructors new_quant_param() and new_qual_param() is deprecated. value_seq() now uses the same logic to generate a sequence of parameter values regardless of how long that sequence is (#153, #229).

• prior_terminal_node_expo() for Bayesian adaptive regression trees (BART) now defaults to a range greater than 1 to limit explosive tree growth (#251).

• The label for spline_degree() was improved.

dials 1.0.0

• The new parameter mtry_prop() is a variation on mtry() where the value is interpreted as the proportion (rather than the count) of predictors that will be randomly sampled at each split (#233).

• conditional_test_statistic() and conditional_test_type() now work with the partykit engine rather than the party engine (#224).

• new_quant_param() no longer requires range and inclusive if values is supplied (#87).

dials 0.1.1

• The Chicago data set was removed. It can be found in the modeldata package.

dials 0.1.0

New parameters

• summary_stat() is a new parameter for use in recipes::step_window().

• A general class_weights() parameter was added for imbalanced models.

• prior_outcome_range(), prior_terminal_node_coef(), and prior_terminal_node_expo() are new parameters for prior distribution specification related to parsnip::bart().

• num_knots() and survival_link() are new parameters for spline survival models (@mattwarkentin, #195).

• vocabulary_size() is a new parameter used in textrecipes::step_tokenize_sentencepiece() and textrecipes::step_tokenize_bpe().

Other changes

• The new extract_parameter_dials() method supersedes pull_dials_object() which has been deprecated.

• activation() now supports values of "tanh".

• New link to article explaining how to make custom parameter objects was added to the pkgdown site.

dials 0.0.10

New parameters

• adjust_deg_free() and select_features() for generalized additive models.

• conditional_min_criterion, conditional_test_statistic, conditional_test_type for models with the party engine.

• diagonal_covariance(), regularization_method(), shrinkage_correlation(), shrinkage_frequencies(), and shrinkage_variance() for linear and quadratic discriminant analysis.

• penalty_L1(), penalty_L2(), and scale_pos_weight() for boosted trees with the xgboost engine (@joeycouse, #176).

• prior_mixture_threshold() and prior_slab_dispersion() for sparse PCA.

• stop_iter() for early stopping.

Other changes

• Re-licensed package from GPL-2 to MIT. See consent from copyright holders here.

• param_set(), scheduled for removal in version 0.0.5, is now removed.

dials 0.0.9

• Quantitative parameters now print the number of possible values if they have been set with value_set() (@kmdupr33, #138).

• The print() method for parameters() has changed to be more clear.

• A new function, pull_dials_object() was also added.

• Duplicate parameter combinations are now automatically removed from grid results.

• The range for epochs() was change to start at 10 iterations instead of 1.

• The lower range for spline_degree() now starts at 1 instead of 3.

• The upper range for cost() now goes to 2^5 instead of 2^-1.

dials 0.0.8

• A number of new parameter objects associated with engine-specific functions were added for engines “ranger”, “randomForest”, “earth” and “C5.0”.

dials 0.0.7

• The grid_*() functions no longer generate subclassed tibbles.

• predictor_prop() was added.

• The levels argument for tune_grid() can now handle a named vector, to account for differences in ordering.

Breaking changes

• The range of dist_power() was changed so that the lower limit is 1.

• The deprecation period for margin() is over; please use svm_margin() instead.

dials 0.0.6

• Quick bug fix release related to range checks in 0.0.5. The check is more forgiving when the required type is integer and a double is provided.

dials 0.0.5

• When kept in the original units, a parameter’s range must now be the same data type as the parameter.

• Renamed margin() to svm_margin() (@gabrielodom and @gralgomez, #85)

• A bug in space filling designs with qualitative parameters was fixed (#94)

• A better error message was created when grids are used with parameters sets that contain parameters that require finalization (#99)

• Space-filling desings now share the same attributes as other grid objects (#100)

• The range for sample_frac() was fixed (#96)

dials 0.0.4

New parameter functions:

• Parameters smoothness() was added.

Other changes

• param_set() is being renamed parameters(). The old name implied that you only use it to set parameters (say from a recipe or model rather than de novo). param_set() will be available until version 0.0.5.

• The range for num_hash() was increase to be 2^8 to 2^12.

• The range for max_tokens() was changed to be 0 to 1000.

Breaking changes

• offset() has been renamed kernel_offset() to avoid name conflicts.

dials 0.0.3

Breaking changes

• All parameter objects are now parameter functions. For example, the pre-configured object mtry is now a function, mtry(), with arguments for the range and the trans. This provides greater flexibility in parameter creation, and should feel more natural.

• deg_free() erroneously produced real values; integers are now returned.

• Default ranges were changed for learn_rate() and neighbors() were changed.

• update.param_set() now takes multiple named arguments.

Other changes

• Two functions for space-filling designs were added: grid_max_entropy() and grid_latin_hypercube().

• A data set was added for modeling ridership on the Chicago L trains.

New parameter functions:

• Parameters spline_degree(), over_ratio(), under_ratio(), freq_cut(), unique_cut(), num_breaks(), min_unique(), num_hash(), signed_hash(), sample_prop(), window_size(), min_dist(), and degree_int() were added.

dials 0.0.2

• Parameter objects now contain code to finalize their values and a number of helper functions for certain data-specific parameters. A force option can be used to avoid updating the values.

• Parameter objects are printed differently inside of tibbles.

• regularization was changed to penalty in a few models to be consistent with this change.

• batch_size and threshold were added.

• Added a set of parameters for the textrecipes package issue 16.

dials 0.0.1

• First CRAN version