- Version
2.2
- Authors
Fernando Rios, Martin Hunt
- Web site
- Documentation
- Copyright
This document has been placed in the public domain.
- License
MIT License.
See https://github.com/martin-hunt/puq for the original readme.
This is a significantly modified version of PUQ by Martin Hunt. It is meant to be used with the Wiggly UA/SA framework. However, it should be mostly backwards compatible with the original PUQ. This fork removes some of the *nix specific system calls which allows PUQ to be run on the Windows platform. Furthermore, some functionality is added.
In addition to running on Windows, this version of puq adds the following functionality:
- Ability to conduct sensitivity analysis using the Morris method (requires my version of the SALib library)
- Can use a Python function as the TestModel. This SIGNIFICATLY speeds up model runs when the model is Python module since a Python interpreter no longer needs to be started for each model run.
- Ability to pass parameters to the test program by file instead of passing them all on the command line.
- Ability to exactly specify the values to be used when conducting an analysis, instead of having puq sample the PDF. This is useful in cases when the samples of the parameter are generated externally.
- Ability to conduct a dry run (all steps of the run are shown, including the command lines to be executed, except the actual model is not run.) In order to have a complete output file, a dummy output value is used.
- Ability to specify the number of samples to use when generating a PDF of a response surface from a script.
The installation procedure and dependencies are the same as https://github.com/martin-hunt/puq. To conduct a Morris sensitivity analysis, the version of SALib located at https://github.com/zoidy/SALib is required. To build sparse_grid_cc you will need a C++ compiler such as gcc. If using MinGW, install (for a 32 bit system),
- mingw32-base (I used 2013072200)
- mingw32-gcc-g++ (I used 4.8.1-4)
and make sure mingw32bin is on your path. Then running the installation:
python setup.pywill cause sparse_grid_cc to build. Afterwards, put this library into the site-packages folder.
Most functionality of the original puq should work. At the moment, only the InteractiveHost is supported.