by Gunnar König, Christoph Molnar, Bernd Bischl, Moritz Grosse-Wentrup [arXiv]
This repository contains the code accompagnying the paper "Relative Feature Importance", submitted at ICPR2020.
- datagen.py: code for the sampling of the dataset used in the examples
- rfi.py: functions for the computation of relative feature importance
- ex1_confounding: code for the example with variables outside training feature set
- ex2_chain: code for indirect influence example
matplotlib==3.2.2
seaborn==0.10.1
torch==1.5.0
pyro_ppl==1.3.1
pandas==1.0.5
numpy==1.18.1
scipy==1.4.1
pyro==3.16
scikit_learn==0.23.1
An updated implementation of RFI can be found in the fippy
package on https://github.com/gcskoenig/fippy.