IRIS-HEP Fellow: Jason Hipkins

Fellowship dates: Jun – Sep, 2021

Home Institution: University of Maryland - College Park

Project: Advancing an active learning algorithm for more efficient generation of Monte Carlo for exclusion plots

Current methods for computing excursion sets of black-box functions (equivalently finding iso-hypersurfaces of n-dimensional scalar multivariate functions) are embarrassingly parallel and computationally expensive. An active learning algorithm appropriately named ‘excursion’ has reduced the compute resources necessary to find excursion sets so that researchers can quickly classify BSM theories. Using a Bayesian optimization procedure it can compute excursion sets in record times. Our goal is to scale this method so that it still works efficiently in higher dimensions. It will be extremely helpful in the search for new physics.

More information: My project proposal

Mentors:

• Lukas Heinrich (CERN)

• Kyle Cranmer (New York University)

• Irina Espejo Morales (New York University)

Presentations and Publications

• 29 Sep 2021 - "Active Learning for Excursion Set Estimation", Jason Hipkins, IRIS-HEP Topical Meetings Recording: Active Learning for Excursion Set Estimation

---

Current Status


Contact me:

• Send me an email
  
• Visit my GitHub Repo
  
• Visit my LinkedIn Profile