SLAC group to focus on analysis with machine learning techniques

Adapted from the SLAC article Hirohisa Tanaka Joins SLAC to Push Limits of Neutrino Physics of 3 April 2018.

Accomplished neutrino physicist Hirohisa Tanaka has joined the Department of Energy’s SLAC National Accelerator Laboratory as a professor of particle physics and astrophysics. He oversees a group at the lab that is preparing for research with the future Deep Underground Neutrino Experiment (DUNE) at the Long-Baseline Neutrino Facility (LBNF).

Members of SLAC’s LBNF/DUNE group. From left: Hirohisa Tanaka (professor), Ji Won Park (rotation student), Tracy Usher (staff scientist), Laura Domine (graduate student), Leon Rochester (staff scientist, retired), Mark Convery (staff scientist), Brandon Eberly (postdoctoral researcher), Kazuhiro Terao (associate staff scientist). Not pictured: Matt Graham (staff scientist), Patrick Tsang (project scientist), Yun-Tse Tsai (associate staff scientist). (Dawn Harmer/SLAC National Accelerator Laboratory)

Tanaka’s appointment is the latest move to expand SLAC’s profile in neutrino research – a scientific area that has become a major focus of U.S. particle physics in recent years.

So far, the SLAC group has been focusing on developing the data processing system for future neutrino experiments. It consists of electronics that turn the raw signal from a detector into a digital one; data acquisition systems to record this data; and software to determine if a signal was caused by a neutrino and to compute the neutrino’s properties, such as type and energy.

The researchers are exploring the possibility of analyzing events recorded by the detector with machine learning techniques inspired by advanced pattern recognition methods developed for computer vision. They also plan on contributing to the design of the near detector – a key component of the DUNE experiment that will study neutrinos near their source, allowing the separation of neutrino oscillation from other possible effects.

DUNE will also be able to capture neutrino signals from the sky that were produced in powerful cosmic events, such as exploding stars, called supernovas. Unlike the neutrino streams from Fermilab, which will arrive at DUNE at precise intervals, astrophysical signals can happen at any time, making them challenging to study.

“We don’t know when certain astrophysical events will happen, so we need to be always ready for them,” Tanaka says. “But we can’t possibly record everything DUNE will see because the data volume would simply be overwhelming. That’s why we’re developing new techniques that’ll make sure we don’t miss those important sporadic events.”

Read the full article.