mlr 2.8: 2016-02-13

  • Feature filter “univariate” had a bad name, was deprecated and is now called “univariate.model.score”. The new one also has better defaults.
  • (generate/plot)PartialPrediction: added new arg “geom” for tile plots
  • small fix for plotBMRSummary
  • the ModelMultiplexer inherits its predict.type from the base learners now
  • check that learners in an ensemble have the same predict.type
  • new function getBMRModels to extract stored models from a benchmark result
  • Fixed a bug where several learners from the LiblineaR package (“classif.LiblineaRL2LogReg”, “classif.LiblineaRL2SVC”, “regr.LiblineaRL2L2SVR”) were calling the wrong value for “type” (0) and thus training the wrong model.
  • Fixed a bug where the resampling objects hout, cv2, cv3, cv5, cv10 were not documented in the ResampleDesc help page
  • regr.xgboost, classif.xgboost: add feval param
  • fixed a bug in irace tuning interface with unamed discrete values
  • Fixed bugs in “jackknife” and “bootstrap” se estimators for regr.randomForest.
  • Added “sd” estimator for regr.randomForest.
  • Fixed a mini bug in ModelMultiplexer where hyperpars that are only needed in predict were not passed down correctly
  • Fixed a bug where the function capLargeValues wasn’t working if you passed a task.
  • capLargeValues now has a new argument “target”, to prevent from capping response values.
  • classif.gbm, regr.gbm: Updated possible ‘distribution’ settings a bit.
  • oversample, undersample, makeOversampleWrapper, makeUndersampleWrapper, makeOverBaggingWrapper: Added arguments to specifically select the sampled class.

API changes

  • listLearners now returns a data frame with properties of the learners if create is false

new functions

  • getBMRModels

removed functions

  • generateROCRCurvesData, plotROCRCurves, plotROCRCurvesGGVIS

new learners

  • classif.randomForestSRCSyn
  • classif.cvglmnet
  • regr.randomForestSRCSyn
  • cluster.dbscan

new measures

  • rsq, arsq, expvar