Introduction to surmise

surmise is a Python package that is designed to provide a surrogate model interface for calibration, uncertainty quantification, and sensitivity analysis.

Dependencies

surmise is build for Python 3.8 or above, with the following dependencies:

numpy>=1.18.3
scipy>=1.7
scikit-learn>=1.2.0

Installation

From the command line, use the following command to install surmise:

pip install surmise

The list of available .whl files can be found under PyPI-wheel. If a wheel file for your preferred platform is not listed, surmise has to be built from source, which requires extra dependencies:

git clone https://github.com/bandframework/surmise/
cd surmise
pip install -r requirements.txt

Note

Direct installation of surmise requires Cython to build C executable. On Windows platform Cython is supported by Microsoft build tools, for which installation is outside pip. See Microsoft build tools for details.

Testing

Testing of surmise can be performed after cloning the repository. The test suite requires the pytest and pytest-cov packages to be installed. These packages can be installed via:

pip install pytest pytest-cov

The test suite can then be run from within the tests/ directory of the source distribution by running:

cd tests
./run-tests.sh

Further options are available for testing. To see a complete list of options, run:

./run-tests.sh -h

Coverage reports are produced under tests/cov_html directory only if all tests are used.

Documentation

The documentation is stored in docs/ and is compiled with the Sphinx Python documentation generator. It is written in the reStructuredText format. The documentation is hosted at Read the Docs.

To compile the documentation, first ensure that Sphinx and its dependencies are installed. To install Sphinx and/or ensure compatibility of dependencies, run make from a terminal within the docs/ directory:

cd docs
make

To generate documentation, run command make html from a terminal within the docs/ directory:

(cd docs)
make html

The HTML files are then stored in docs/_build/html.

Citation:

Please use the following to cite surmise in a publication:

@techreport{surmise2023,
  author      = {Matthew Plumlee and \"Ozge S\"urer and Stefan M. Wild and Moses Y-H. Chan},
  title       = {{surmise 0.2.1} Users Manual},
  institution = {NAISE},
  number      = {Version 0.2.1},
  year        = {2023},
  url         = {https://surmise.readthedocs.io}
}

Examples

We provide examples in the examples/ directory to illustrate the basic usage of surmise.

In addition, for a gentle introduction of emulation and calibration using Gaussian processes, visit surmise Jupyter notebook.