Facing the calculational bottleneck in high-energy physics

Contact person: Are Rachl?w, Anders Kvellestad
Keywords: High-energy physics, Computational physics, Machine learning, Statistical inference    
Research group: Theoretical Physics
Department of Physics    
 

The search for new physics in modern high-energy experiments faces serious calculational bottlenecks. Comparing theoretical predictions in new physics models to data requires very computationally expensive precision calculations in quantum field theory, large scale event simulation and reconstruction, as well as heavily resource consuming statistical inference. Today, these calculations are too expensive to perform at the necessary precision except for in the simplest of models, severely limiting the possibility of discoveries despite the huge investments in particle physics experiments such as the LHC. Proposals for this theme can span across any number of methodological approaches to unclogging the bottleneck taken from computational science, mathematics and statistics. 

Applications will be mainly within high-energy physics under the auspices of the GAMBIT Collaboration, with the possibility of participating in existing cross-disciplinary collaborations with the nuclear physics section and the Norwegian Institute of Public Health on common methodology, as well as secondments to industry partners.

Methodological research topics:

  • Machine learning regression: development of techniques to speed up quantum field theory (QFT) calculations
  • Continual learning: on-the-fly statistical learning for global fits of new physics models
  • Automation of precision QFT calculations
  • Parallelisation of central high-energy physics algorithms, e.g. jet algorithms, phase-space generation, or parton distribution functions
  • Development of new statistical inference methodology, e.g. for goodness-of-fit of new physics models, or the statistical interpretation of higher-order QFT uncertainties

Note: This list should not be seen as excluding other possible methodological research topics.

External partners:

  • The GAMBIT Collaboration (gambitbsm.org)
  • Norwegian Institute of Public Health
  • Telenor Research

Mentoring and internship will be offered by a relevant external partner.