What is Fit-Benchmarking?
FitBenchmarking is an open source Python package that was first developed to compare how different fit minimizers perform on various fitting problems. The original version was designed to work the widely adopted Neutron data reduction software Mantid.
Useful skills to have
- A strong desire to work on some challenging software issues as part of a team
- Knowledge of Python and optimisation
- Previously worked with git or similar version control
About the Project
Instrument scientists at ISIS spend a large amount of their time fitting (optimising) models against data. The 'engines' that do this are called minimizers. Existing neutron software that depends on fitting favours a spectrum of minimizers and/or different implementations of the same minimizers. The FitBenchmarking tool is the first tool that aims to compare how minimizers, as they are implemented in such software, perform against fit problems commonly executed.
The goal of this exciting summer project is to further develop the FitBenchmarking tool. Specifically, to extend FitBenchmarking so that it can use minimizers from the world leading small angle scattering software SasView and to extend the fitting problems to include data from SasView. This will then also allow a direct comparison between SasView's and Mantid's minimizer implementations against a wide range of fitting problems.
The successful applicant will be based in the Analysis, Analytics and Infrastructure section of ISIS, and will work closely with Computational Mathematics Group of SCD, SasView and Mantid team members.
Benefits to the Student
- Improved programming and software engineering skills
- Working closely with SasView scientists, Computational Mathematicians and Software engineers
- Being part of a global team of scientists and software developers
- An enhanced understanding of optimisation and modelling
Supervisor: Anders Markvardsen, email@example.com