A Voronoi cell interface-based parameter sensitivity analysis technique. Reference implementation for the paper Voronoi Cell Interface-Based Parameter Sensitivity Analysis for Labeled Samples.

Tested with Python 3.12.3.

1. Setup virtual environment with Python: python -m venv .venv
2. Activate the environment, then install the requirements via: pip install -r requirements.txt
3. Run python ./main.py to run the main example.

Some of the figures contained in the Jupyter Notebooks were used to generate the CCY BY 4.0 licensed figures in https://doi.org/10.1111/cgf.70122. The figures here are the raw .png figures. The published figures are mostly vector graphics and were slightly modified compared to the figures rendered in the Jupyter notebooks.

• Author: Ruben Bauer
• Source: https://doi.org/10.1111/cgf.70122
• License: CCY BY 4.0

The figures of the main paper are:

• examples/notebooks/plot_examples/two_d_clipping_and_distribution.ipynb used to generate Figure 2, 4
• examples/notebooks/plot_examples/three_dimensional_case.ipynb used to generate Figure 3
• examples/notebooks/plot_examples/two_d_major_transition_directions.ipynb used to generate Figure 5
• examples/notebooks/single_dataset_convenience_plots/iris_conv_plot.ipynb used to generate Figure 6, 7, 8, 9, 10, 12
• examples/notebooks/plot_examples/space_dividing_line.ipynb used to generate Figure 11
• examples/notebooks/single_dataset_convenience_plots/semiconductor_conv_plot.ipynb used to generate Figure 14
• examples/notebooks/single_dataset_convenience_plots/droplet_impact_conv_plot.ipynb used to generate Figure 15, 16

For the supplemental material:

• examples/notebooks/plot_examples/normal_vectors_angle_vis.ipynb used to generate Figure 1
• examples/notebooks/plot_examples/plane_plane_distance.ipynb used to generate Figure 2
• examples/notebooks/plot_examples/runtime_experiments_2.ipynb used to generate Figure 3, 4
• examples/notebooks/plot_examples/bandwidth_experiments.ipynb used to generate Figure 5 - 15

This package uses a fork of the pyclustering library with a customized cluster update method. Do not use that version for regular k-Means but refer to the original pyclustering implementation instead.

• To run the Python files example in the ./examples/ directory, go to the directory via cd examples and then run python ./the_example_to_run.py
• To run the Jupyter Notebooks in the ./examples/notebooks/ directory, select the virtual environment as Python Interpreter first, then run them.

@article{https://doi.org/10.1111/cgf.70122,
  author = {Bauer, R. and Evers, M. and Ngo, Q. Q. and Reina, G. and Frey, S. and Sedlmair, M.},
  title = {Voronoi Cell Interface-Based Parameter Sensitivity Analysis for Labeled Samples},
  journal = {Computer Graphics Forum},
  pages = {e70122},
  keywords = {CCS Concepts, • Human-centered computing → Information visualization, Visual analytics},
  doi = {https://doi.org/10.1111/cgf.70122},
  url = {https://onlinelibrary.wiley.com/doi/abs/10.1111/cgf.70122},
  eprint = {https://onlinelibrary.wiley.com/doi/pdf/10.1111/cgf.70122},
}