Neural Network by Ear: How to Capture a Particle's Energy express
Measuring a particle’s energy at the Large Hadron Collider is like trying to catch a whisper in a roaring crowd. When protons collide, a flurry of particles is born, their signals blending into a cacophony. Old-school methods, like a noise-clogged ear, systematically underestimate the energy.
Scientists built a neural network that works right inside the detector. It tunes into the din and picks out the faint voice of the wanted particle. The top versions of the network are off by mere fractions of a percent — like mistaking 80 grains of sand out of a million. And while the network was trained on simulated events, it flawlessly recognizes real data.
Now physicists can study elusive particles, like dark matter, more precisely, even in heavy noise. And all this insight fits on a chip smaller than a matchbox.
Original abstract
as written by the authors arXiv:2510.11469
A study of neural network architectures for the reconstruction of the energy deposited in the cells of the ATLAS liquid-argon calorimeters under high pile-up conditions expected at the HL-LHC is presented. These networks are designed to run on the FPGA-based readout hardware of the calorimeters under strict size and latency constraints. Several architectures, including Dense, Recurrent (RNN), and Convolutional (CNN) neural networks, are optimised using a Bayesian procedure that balances energy resolution against network size. The optimised Dense, CNN, and combined Dense+RNN architectures achieve a transverse energy resolution of approximately 80 MeV, outperforming both the optimal filtering (OF) method currently in use and RNNs of similar complexity. A detailed comparison across the full dynamic range shows that Dense, CNN, and Dense+RNN accurately reproduce the energy scale, while OF and RNNs underestimate the energy. Deep Evidential Regression is implemented within the Dense architecture to address the need for reliable per-event energy uncertainties. This approach provides predictive uncertainty estimates with minimal increase in network size. The predicted uncertainty is found to be consistent, on average, with the difference between the true deposited energy and the predicted energy.
The ATLAS detector is as big as an eight-story building, but the neural network processing its signals fits on a chip smaller than a matchbox.