TY  - JOUR
PB  - SciPost Foundation
DO  - 10.21468/SciPostPhysCore.8.4.073
TI  - Accurate surrogate amplitudes with calibrated uncertainties
PY  - 2025/10/24
UR  - https://www.scipost.org/SciPostPhysCore.8.4.073
JF  - SciPost Physics Core
JA  - SciPost Phys. Core
VL  - 8
IS  - 4
SP  - 073
A1  - Bahl, Henning
AU  - Elmer, Nina
AU  - Favaro, Luigi
AU  - Haussmann, Manuel
AU  - Plehn, Tilman
AU  - Winterhalder, Ramon
AB  - Neural networks for LHC physics have to be accurate, reliable, and controlled. Using neural surrogates for the prediction of loop amplitudes as a use case, we first show how activation functions are systematically tested with Kolmogorov-Arnold Networks. Then, we train neural surrogates to simultaneously predict the target amplitude and an uncertainty for the prediction. We disentangle systematic uncertainties, learned by a well-defined likelihood loss, from statistical uncertainties, which require the introduction of Bayesian neural networks or repulsive ensembles. We test the coverage of the learned uncertainties using pull distributions to quantify the calibration of cutting-edge neural surrogates.
ER  -

@Article{10.21468/SciPostPhysCore.8.4.073,
	title={{Accurate surrogate amplitudes with calibrated uncertainties}},
	author={Henning Bahl and Nina Elmer and Luigi Favaro and Manuel Haußmann and Tilman Plehn and Ramon Winterhalder},
	journal={SciPost Phys. Core},
	volume={8},
	pages={073},
	year={2025},
	publisher={SciPost},
	doi={10.21468/SciPostPhysCore.8.4.073},
	url={https://scipost.org/10.21468/SciPostPhysCore.8.4.073},
}