Negative weight mitigation with cell resampling in ATLAS workflows
Description
Negatively weighted events arise as a result of subtraction schemes in next-to-leading (or higher) order event generators. The fraction of negatively weighted events vary as a function of phase space requirements that are imposed in experimental analyses making it imperative to store these events for time consuming downstream processing like detector simulation. They are a severe source of inefficiency in event generator workflows, requiring large datasets to mitigate statistical dilution caused by negatively weighted events.
A method to redistribute negatively weighted events was proposed in arXiv:2109.07851 and subsequently in arXiv:2303.15246. We plan to use this method for ATLAS event generator workflows. The method has been previously implemented in CMS for small-scale tests. In this project, we will extend the scope of previous explorations in both ATLAS and CMS by identifying computationally intensive workflows and running validation tests that are designed to ensure that distributions of physical observables are not sculpted as a result of the removal of negatively weighted events.
The eventual goal of the project is to integrate the negative weight mitigation scheme into a realistic ATLAS workflow and setup a validation pipeline that ensures that the method is performing as desired.
Task ideas
- Establish familiarity with ATLAS event generator workflows
- Run cell resampling method with fake data (generated with a pseudorandom generator thrown from distributions that are indicative of experimental data)
- Run cell resampling with ATLAS event generator workflows
- Setup a validation suite
- Document results with distributions of variables before and after the method has been applied with a metric that shows computational gains in terms of lower fraction of negatively weighted events
Expected results
- Design an event generator workflow and validation suite that tests the cell resampling method for negative weight removal
Requirements
- Familiarity with Python and C++
- Interest in learning Rust
Links
AI usage policy
AI assistance is allowed for this contribution. The applicant takes full responsibility for all code and results, disclosing AI use for non-routine tasks (algorithm design, architecture, complex problem-solving). Routine tasks (grammar, formatting, style) do not require disclosure.
Mentors
- Saptaparna Bhattacharya - SMU
- Jeppe Andersen - IPPP
- Andreas Maier - DESY
Additional Information
- Difficulty level (low / medium / high): medium
- Duration: 175 hours
- Mentor availability: June-October