Skip to content

Releases: thomasp85/lime

lime 0.5.3

19 Aug 08:50
Compare
Choose a tag to compare
  • Emil Hvitfelt is taking over maintenance
  • General upkeep

lime 0.5.2

24 Feb 15:25
Compare
Choose a tag to compare
  • Fixed use of order() on data.frame objects
  • Moved htmlwidgets, shiny, and shinythemes to suggests

lime 0.5.1

12 Nov 08:21
49df0a1
Compare
Choose a tag to compare
  • Fixed namespace import from glmnet following changes there

lime 0.5.0

24 Jun 11:11
e86cb10
Compare
Choose a tag to compare
  • explain() will now pass ... on to the relevant predict() method (#150)
  • explain.data.frame() gains a gower_pow argument to modify the calculated
    gower distance before use by raising it to the power of the given value (#158)
  • Fixed a bug when calculating R^2 on single feature explanations (@pkopper, #157)
  • Fixed formatting of text prediction html presentation (#145)
  • Fixed a bug when setting feature select method to "none" (#141)
  • Changes default colouring from green-red to blue-red (#137)
  • lime() now warns when quantile binning is not feasible and uses standard
    binning instead (#154)
  • Changed the lambda value in the local model fit to match the one used in the
    Python version according to the relationship given here:
    https://stats.stackexchange.com/a/270705
  • Added pkgdown site at https://lime.data-imaginist.com
  • Fixed a bug when using a proprocessor with data.frame explanations

lime v0.3.1

24 Nov 10:36
lime-0.3.1
537ffa7
Compare
Choose a tag to compare
  • Added a NEWS.md file to track changes to the package.
  • Fixed bug when explaining regression models, due to drop=TRUE defaults (#33)
  • Integer features are no longer converted to numeric during permutations (#32)
  • Fix bug when working with xgboost and tabular predictions (@martinju #1)
  • Training data can now contain NA values (#8)
  • Keep ordering when plotting with plot_features() (#38)
  • Fix support for mlr by extracting predictions correctly
  • Added support for h2o (@mdancho84) (#40)
  • Throws meaningful error when all permutations have 0 similarity to original
    observation (#47)
  • Explaining data can now contain NA values (#45)
  • Support for Date and POSIXt columns. They will be kept constant during
    permutations so that lime will explain the model behaviour at the given
    timepoint based on the remaining features (#39).
  • Add plot_explanations() for an overview plot of a large explanation set