Movatterモバイル変換

ingredients 2.3.0

• breaking change:calculate_variable_splits() now treatsinteger variables ascategorical. This changeis propagated toceteris_paribus(),partial_dependence(),accumulated_dependence(),conditional_dependence(),aggregate_profiles(),DALEX::predict_profile(),DALEX::model_profile()
• fix an error inceteris_paribus /calculate_variable_splits whentidymodels usesinteger variables#145
• fix an error inshow_observations#148.This change is propagated toDALEX::plot.predict_profile()#540.
• fix#149by replacing allclass(x) = "y" withis(x, "y")

ingredients 2.2.1

• addedfacet_scales parameter toplot.aggregated_profiles_explainer ('free_x'by default)#138andplot.ceteris_paribus_explainer ('free_x'or'free_y' by default, depending on plot type)#136

ingredients 2.2.0

• fixes explanations when data has one column#137

ingredients 2.0.1

• code and documentation maintenance#130
• fixed an error whenN = NULL inpartial_dependence() etc.#134

ingredients 2.0

• plot.ceteris_paribus_explainer now by default forcategorical variables plots profiles (not lines -prev default- norbars)
• ALE plots are now centered around average y_hat#126
• colors from DrWhy color palette is used for CP#125

ingredients 1.3.1

• defaultsubtitle value inplot.fi changedtoNULL fromNA (unification)
• now in theceteris_paribus function one can specify howgrid points shall be calculated, seevariable_splits_type
• ceteris_paribus and aggregates are now working withmissing data, this solves#120
• plot(ceteris_paribus) change defaultcolortolabel orids if more than one profile is detected,this solves#123
• ceteris_paribus has now argumentvariable_splits_with_obs which included values fromnew_observations in thevariable_splits, thissolves#124

ingredients 1.3.0

• deprecaten_sample argument infeature_importance (now it’sN)#113
• plot_profile now handles multilabel models

ingredients 1.2.0

• DALEX is moved to Suggests as in#112
• plot_categorical_ceteris_paribus can plot bars(again)
• addbind_plots function

ingredients 1.1.0

• supportR v4.0 and depend onR v3.5 tocomply withDALEX
• new argumentstitle andsubtitle inseveral plots

ingredients 1.0.0

• changedependency todependence#103

ingredients 0.5.2

• ceteris_paribus profiles are now working forcategorical variables
• show_profiles,show_observations,show_residuals are now working for categoricalvariables

ingredients 0.5.1

• synchronisation with changes in DALEX 0.5
• new argumentdesc_sorting inplot.variable_importance_explainer#94

ingredients 0.5.0

• feature_importance now does15permutations on each variable by default. Use theBargument to change this number
• added boxplots toplot.feature_importance andplotD3.feature_importance that showcase the permutationdata
• inaggregate_profiles: preserve_x_ columnfactor order and sort its values#82

ingredients 0.4.2

• aggregate_profiles use now gaussian kernel smoothing.Use thespan argument for fine control over this parameter(#79)
• changevariable_type andvariablesarguments usage in theaggregate_profiles,plot.ceteris_paribus andplotD3.ceteris_paribus
• removevariable_type argument fromplotD3.aggregated_profiles (now the same as inplot.aggregated_profiles)
• Kasia Pekala is moved as contributor to theDALEXtra asaspect_importance is moved toDALEXtra as well(Seev0.3.12 changelog)
• added Travis-CI for OSX

ingredients 0.4.1

• fixed rounding problem in the describe function (#76)

ingredients 0.4

• CRAN release

ingredients 0.3.12

• aspect_importance is moved toDALEXtra (#66)
• examples are updated in order to reflect changes intitanic_imputed fromDALEX (#65)

ingredients 0.3.11

• modifiedplot.aspect_importance - it can plot more thansingle figure
  
• modifiedtriplot,plot.aspect_importanceandplot_group_variables to add more clarity in plots andallow some parameterization

ingredients 0.3.10

• addedtriplot function that illustrates hierarchicalaspect_importance() groupings
• changes inaspect_importance() functions
• added back the vigniette foraspect_importance()

ingredients 0.3.9

• changeonly_numerical parameter tovariable_type in functions aggregated_profiles(),cluster_profiles(), plot() and others, as requested in #15

ingredients 0.3.8

• Natural language description generated withdescribe()function forceteris_paribus(),feature_importance() andaggregate_profiles()explanations.

ingredients 0.3.7

• aggregated_profiles_conditional andaggregated_profiles_accumulated are rewritten with somecode fixes

ingredients 0.3.6

• a new version oflime is implemented in thelime()/aspect_importance() function.
• Kasia Pekala and Huber Baniecki are added as contributors.

ingredients 0.3.5

• new feature#29.Feature importance now takes an argumentB that replicatespermutationsB times and calculates average from droploss.

ingredients 0.3.4

• plotD3 now supports Ceteris Paribus Profiles.
• feature_importance now can takevariable_grouping argument that assess importance of groupof features
• fix in ceteris_paribus, now it handles models with just onevariable
• fix#27for multiple rows

ingredients 0.3.3

• show_profiles andshow_residuals functionsextend Ceteris Paribus Plots.
• show_aggreagated_profiles is renamed toshow_aggregated_profiles
• centering of ggplot2 title

ingredients 0.3.2

• added new functionsdescribe() andprint.ceteris_paribus_descriptions() for text baseddescriptions of Ceteris Paribus explainers
• plot.ceteris_paribus_explainer works now also forcategorical variables. Use theonly_numerical = FALSE toforce bars

ingredients 0.3.1

• added references to PM VEE
• partial_profiles(),accumulated_profiles()andconditional_profiles for variable effects
• major changes in function names and file names

ingredients 0.3

• ceteris_paribus_2d extends classical ceteris paribusprofiles
• ceteris_paribus_oscillations calculates oscilations forceteris paribus profiles
• fixed examples and file names

ingredients 0.2

• cluster_profiles helps to identify interactions
• partial_dependency calculates partial dependencyplots
• aggregate_profiles calculates partial dependency plotsand much more

ingredients 0.1

• port ofmodel_feature_importance andmodel_feature_response fromDALEX toingredients
• added tests