Deep Learning at the Frontier of Particle Physics
| Date/Time: | Monday, 10 Sep 2018 from 4:10 pm to 5:00 pm |
|---|---|
| Location: | Phys 0003 |
| Contact: | |
| Phone: | 515-294-5441 |
| Channel: | College of Liberal Arts and Sciences |
| Actions: | Download iCal/vCal | Email Reminder |
Abstract: Our knowledge of the fundamental particles of nature and their interactions is elegantly summarized by the standard model of particle physics. Advancing our understanding in this field has required experiments that operate at even higher energies and intensities, which produce extremely large and information-rich data samples. The use of advanced machine-learning techniques is revolutionizing how we interpret these data samples, greatly increasing the discovery potential of present and future experiments.
This talk will summarize the challenges and opportunities that come with the use of deep learning at the frontiers of particle physics, with a particular focus on examples from the thriving subfield of neutrino physics.
Bio: Since graduate school, I have made essential contributions to the study of neutrino oscillations through key work on the MINOS, NOvA, and DUNE experiments. I have worked on specific oscillation analyses, such as muon neutrino disappearance on MINOS and electron neutrino appearance on NOvA, and developed general calibration, neutrino beam simulation, and particle identification tools which have had cross cutting benefits throughout those experiments. I now lead the muon neutrino disappearance effort on NOvA as co-convener on the working group, directing a major update to the neutrino result released January 2018, and a first measurement using an anti-neutrino beam released in June 2018.
As a postdoc, I have been heavily involved in the first applications of deep learning in the field of particle physics. Most notably, I oversaw the first use of a CNN in published particle physics analysis, which dramatically improved the power of the NOvA experiment's headline analyses. I am a founding member of the Fermilab-based Machine Learning group, an organization which aims to foster inter-experiment collaboration on machine learning tools. On DUNE I've developed the only particle ID which gives adequate performance- hugely outperforming more traditional approaches and already slightly outperforming the estimates of human hand scanners from the DUNE CDR. Most recently, I am lead author on a review of machine learning in particle physics, which was published in Nature this August. This review attempts to cover the unique role and challenge of machine learning at both the intensity and energy frontiers of particle physics.
-- PhD at UCL with Jenny Thomas, updated the MINOS oscillation analysis to a full three neutrino flavor framework.
-- Postdoc on NOvA/MINOS/DUNE with William and Mary for last five years.
-- Convened NOvA and MINOS beam groups, guiding simulation of the NuMI neutrino beam.
-- Developed novel ways to quantify uncertainties on the simulation, essential for the MINOS sterile neutrino analysis.
-- Key member of the NOvA electron neutrino appearance group through it's first two analyses.
-- Initiated, designed, and coordinated resources for the NOvA experiment's first deep learning network, which improved the NOvA headline measurement by 30%. Published the algorithm and application in the Journal of Instrumentation (arXiv:1604.01444).
-- Convened the NOvA muon neutrino disappearance group, guiding it through two major result releases, including both a significant analysis update and the first analysis of anti-neutrino data at NOvA. --Directed work on the DUNE Deep Learning based particle ID, the only ID to achieve the design performance of DUNE in current simulation.
-- Lead author on a review of machine learning in particle physics, published in Nature (doi:10.1038/s41586-018-0361-2).