/iPDP-On-Partial-Dependence-Plots-in-Dynamic-Modeling-Scenarios

Technical Supplement of Research Work

Primary LanguagePythonMIT LicenseMIT

Paper Supplement: iPDP: On Partial Dependence Plots in Dynamic Modeling Scenarios

The experiments and results presented in the paper are available in the projects/iPDP module. The projects/iPDP module contains the following submodules:

  • batch_california.py: Script to run the experiments on the California Housing dataset.
  • change_detector_elec2.py: Script to run the experiments on the Elec2 data stream.
  • agrawal_concept_drifts.py: Create and run the experiment on the Agrawal data stream.
  • change_detector.py: Module containing the change detector class.
  • rotating_plane.py: Script to run the experiments on the Rotating Hyperplane data stream.

The implementation of the iPDP can be found in the iXAI module. Specifically, the iXAI.ixai.explainers.pdp module contains the implementation of the iPDP. The iXAI.ixai.utils.tracker.extreme_value_tracker module contains the implementation of the maximum/minimum value tracker. The iXAI/ixai/storage/ordered_reservoir_storage contains the extended reservoir storage implementation for frequency storage mechanisms.

Running the experiments

  1. Install requirements by running pip install -r requirements.txt.
  2. Navigate to iXAI root directory (cd iXAI) and install the ixai package by running pip install ..
  3. Navigate to the projects/iPDP directory (cd projects/iPDP) and run the experiments.