Description
The knowledge of the proton structure, embedded in the parton distribution functions (PDFs) is of fundamental importance to make predictions for proton-proton collisions at the Large Hadron Collider. PDFs are determined through fits to experimental data, and the functional form assumed plays an important role. Using too few PDF parameters, or a constrained parametrisation, can lead to artificially small PDF uncertainties in certain regions of the phase-space, or to difficulties in minimising the PDF parameters.
Aim of this project is to interface the Neural Networks Analytic Derivatives (NNAD) package to the xFitter code. This will allow for a PDF parametrisation based on NN, thus with minimal assumptions on their functional behavior, and, thanks to the implementation of analytic derivatives, for a fast and efficient minimisation using the CERES non-linear least squares solver. The NN parametrisation will be tested using an ongoing extraction of PDFs using measurements from the CMS experiment.
Special Qualifications:
- knowledge of C++
| Field | B6: Computing |
|---|---|
| DESY Place | Hamburg |
| DESY Division | FH |
| DESY Group | CMS |
