ModelOriented / ingredients

Effects and Importances of Model Ingredients
https://modeloriented.github.io/ingredients/
GNU General Public License v3.0
37 stars 19 forks source link

Support 3d arrays as input data for feature_importance #141

Open jmaspons opened 2 years ago

jmaspons commented 2 years ago

Useful for RNN such as LSTM or GRU. All tests from test_variable_dropout.R pass

It implements permutations for the 2nd and/or 3rd dimensions (eg. for a time series, permute cases and time steps for the same var instead of permuting cases only, or considering the importance of each combination of time step and feature)

I don't know if it's OK to save an example model for testing or if it's ok to run a keras model with dummy data on tests. Tell me what do you prefer