llnl / ddd

Delaunay density diagnostic

Delaunay Density Diagnostic

Version 2.0, January 2024.

This code implements algorithms described in:
Algorithm XXXX: The Delaunay Density Diagnostic
under review at ACM Transactions on Mathematical Software
original title: ``Data-driven geometric scale detection via Delaunay interpolation'' Andrew Gillette and Eugene Kur, 2022
https://arxiv.org/abs/2203.05685

Usage

1. Activate a python environment that includes the packages listed in the REQUIREMENTS.txt file.

2. Ensure that the gfortran compiler is installed.

3. Run the driver script for the Griewank and/or static data examples:

   python run_ddd_griewank.py
   

   The above script will run a total of 100 trials of the delaunay_density_diagnostic.py script, using data from the 2D Griewank function. The results are saved as .csv files. Then the script generate_ddd_figures.py is called to generate a .png figure called ddd-figure-griewank.png. The figure should match the file ddd-figure-griewank-repo.png that is contained in the repository. More details can be found in the header of run_ddd_griewank.py.

   A typical run time for a single trial is a few seconds, so the whole script should complete in 5-10 minutes.

   python run_ddd_static.py
   

   The above script will run a total of 100 trials of the delaunay_density_diagnostic.py script, using data from the static topography dataset described in the paper (and stored in the subfolder staticdata/). The results are saved as .csv files. Then the script generate_ddd_figures.py is called to generate a .png figure called ddd-figure-static.png. The figure should match the file ddd-figure-static-repo.png that is contained in the repository. More details can be found in the header of run_ddd_static.py.

4. If the figures generates correctly, run

   python delaunay_density_diagnostic.py --help
   

   to see the command line options that can be added to the driver scripts for user-specified experiments.

Debugging notes

The package includes source files in Fortran that impmlement a version of TOMS Algorithm 1012: DELAUNAYSPARSE. This version that has been updated from the original submission to more easily allow python wrapping. Running the script delaunay_density_diagnostic.py will compile the relevant files using gfortran.

During compiling, this type of warning may occur:

Warning: Rank mismatch between actual argument at (1) and actual argument at (2)

This warning is issued by the slatec library that is included with the DELAUNAYSPARSE source code and is not easily suppressed. However, this warning is only due to a change in Fortran conventions since the original publication of TOMS 1012 and does not cause any issues in regards to the results.

Authors

The Delaunay density diagnostic code was created by Andrew Gillette, [email protected], with input from Eugene Kur, [email protected].

Citation information

If you are referring to this code in a publication, please cite the following paper:

Andrew Gillette and Eugene Kur. Data-driven geometric scale detection via Delaunay interpolation. Submitted. 2022. LLNL-JRNL-832470.

@article{GK2022,
  author = Gillette, Andrew and Kur, Eugene},
  title = {Data-driven geometric scale detection via Delaunay interpolation},
  journal = {Submitted. Preprint at arXiv:2203.05685},
  year = {2022},
}

If you wish to cite the code specifically, please use:

@misc{ doecode_72093,
title = {Delaunay density diagnostic},
author = {Gillette, Andrew K.},
url = {https://doi.org/10.11578/dc.20220324.3},
howpublished = {[Computer Software] \url{https://doi.org/10.11578/dc.20220324.3}},
year = {2022},
month = {mar}
}

The DOI for this repository is: https://doi.org/10.11578/dc.20220324.3

License

Delaunay density diagnostic is distributed under the terms of the MIT license.

All new contributions must be made under the MIT license.

See LICENSE and NOTICE for details.

SPDX-License-Identifier: (MIT)

LLNL-CODE-833036