Evolutionary History Matching
This is a Python package which implements History Matching - a statistical analysis technique that calibrates models to data whilst accounting for uncertainty. Currently this is in development and only available as a source installation.
If you would like to test this out, please clone and use the srcinstall.bat script to set everything up. Linux/Bash users will need to adjust this script as it was written for the Windows CMD / PowerShell terminals. Please contact cf502@exeter.ac.uk if you are cloning / forking this project as things are changing day-by-day until there is an official release - it is extremely likely that you will be running old code. At the moment, this repo is collecting snippets of older code that have passed various tests, the full API is being stress-tested before it goes anywhere near this.
A list of features that are in the process of being uploaded slowly is as follows:
- n-Dimensional generic / default History matching (I.E. run as a black-box, will do something 'acceptable' with the right inputs)
- Updated Gaussian Process plugins to support
mogpas well asgpflow - A more robust diagnostic / constraint API to critique emulation
- NROY plotting engine
- Accelerated low dimensional cases where Rejection Sampling can be eliminated for something much better
- Evolutionary ladder methods
- Full save/load/pause/restart controls to facilitate long-running / shared analyses