NEWS.md

In dynfeature: Feature Importance for Dynamic Processes

dynfeature 1.0.1

• MINOR CHANGE: Fixes to the description.

dynfeature 1.0.0

Initial release of dynfeature on CRAN.

• MINOR CHANGE: Use only one core by default.

• MINOR CHANGE: Support sparse matrices.

• DOCUMENTATION: Add examples and returns.

dynfeature 0.2.0 (25-10-2018)

• SPEED UP: Added fi_ranger_rf_lite(), which scales much better w.r.t. the number of samples and features, at the cost of increasing loss of accuracy at higher dimension sizes.

• MAJOR CHANGES: Large cleanup of the code. Most notably,

• The format of feature importance method specification and its parameters, with format fi_method = fi_example_method(param1 = 10, param2 = 4). Before, it had to be specified as method = "example_method", method_params = list(param1 = 10, param2 = 4).

• MINOR CHANGE: Whenever possible, output columns are now factors instead of characters.

• MINOR CHANGE: Add NEWS, and add news section to README.

• DOCUMENTATION: Turned on markdown for Roxygen.

• DOCUMENTATION: Improved documentation on expression_source.

• TESTING: Improved testing with a larger dataset, and will check whether the overall feature importance produces decent results.

• MINOR CHANGE: Feature importance functions will always return factors instead of characters.

dynfeature 0.1.0 (26-04-2018)

• INITIAL RELEASE: dynfeature, calculating feature importance scores from trajectories.

dynfeature documentation built on March 7, 2023, 8:25 p.m.