2 Sources
[1]
Self-learning AI separates particle jets from background noise in simulated heavy-ion collisions
A team of physicists and computer scientists led by researchers at the U.S. Department of Energy's (DOE) Brookhaven National Laboratory developed a new artificial intelligence (AI) approach that learns on its own to separate important high-energy particle jet signals from the enormous background of
[2]
Super Intelligence Method Learns on Its Own to Single Out Jet Signals in Heavy-Ion Collision Data
Newswise -- UPTON, N.Y. -- A team of physicists and computer scientists led by researchers at the U.S. Department of Energy's (DOE) Brookhaven National Laboratory developed a new Super Intelligence (SI) approach that learns on its own to separate important high energy particle jet signals from the
Share
Copy Link
A Brookhaven-led team says a self-learning AI can separate particle jets from background noise without labeled examples. It works in simulation so far; real data is next.
A team led by researchers at Brookhaven National Laboratory has built an AI method that pulls particle jet signals out of the heavy background noise in collision data, and it taught itself to do it without a single prelabeled example
2
1
. The work, published in Scientific Reports with collaborators from Columbia University and the University of Colorado Boulder, aims at the quark-gluon plasma, the hot, dense state of matter that filled the universe a microsecond after the Big Bang; cleaner jet signals could sharpen what physicists learn from it2
1
.Jets are tight sprays of high-energy particles. Physicists consider them one of the best probes of that plasma: as jets cross it, their particles lose energy and their internal structure changes, which reveals the properties of the plasma and of the strong force that governs it
2
1
. The problem is reading them. Background separation has long limited how precisely physicists can measure jets in heavy-ion collisions2
1
.
Source: Phys.org
"The detector sees a crowded 'image' in which the signal we care about is buried inside a dense spray of other particles," said Yeonju Go, a Brookhaven physicist and Stony Brook University assistant professor who led the work
2
.The team claims the method is the first use of an unpaired, unsupervised generative model in this kind of physics, one that learned directly from simulated collision data rather than from human-supplied labels
2
1
. The point of skipping the labels is practical: in real collision data there are no answer-key examples to train against, and building them by hand is slow2
1
."The new framework doesn't need to be told what corresponds to a cat or a dog," said Dmitrii Torbunov, a Brookhaven computer scientist on the project. "Unsupervised learning methods can learn useful concepts and patterns directly from the data, without requiring humans to label every example"
2
.In simulated data from sPHENIX, a detector that ran during the final three runs of Brookhaven's Relativistic Heavy Ion Collider, the model reconstructed jets' positions, energies and internal structure more accurately than conventional background-subtraction methods, with the advantage strongest for larger jets, where background fluctuations bite hardest
2
1
. Better background subtraction, Go said, could lead to sharper measurements of how quarks and gluons behave under extreme conditions2
.There is a catch worth stating plainly: everything reported so far comes from simulated data, and the model has not yet touched real sPHENIX measurements
2
1
. The collider finished operating in early 2026, leaving sPHENIX's nearly 300 petabytes of data awaiting analysis2
1
. The sPHENIX dataset is large enough that extracting physics results from it is the obvious next move, Go said2
1
. Because the framework works directly on image-like detector data, the researchers say it could extend to other experiments trying to pull overlapping signals apart, possibly including the future Electron-Ion Collider2
1
."We definitely want to use the model on our experimental data, especially as sPHENIX has a very large data set," Go said
2
.Related Stories
Summarized by
Navi
[1]
14 Nov 2024•Science and Research
06 Dec 2024•Science and Research

14 Aug 2025•Science and Research
