Bias and Priors in Machine Learning Calibrations for High Energy Physics

Abstract

Machine learning offers an exciting opportunity to improve the calibration of nearly all reconstructed objects in high-energy physics detectors. However, machine learning approaches often depend on the spectra of examples used during training, an issue known as prior dependence. This is an undesirable property of a calibration, which needs to be applicable in a variety of environments. The purpose of this paper is to explicitly highlight the prior dependence of some machine learning-based calibration strategies. We demonstrate how some recent proposals for both simulation-based and data-based calibrations inherit properties of the sample used for training, which can result in biases for downstream analyses. In the case of simulation-based calibration, we argue that our recently proposed Gaussian Ansatz approach can avoid some of the pitfalls of prior dependence, whereas prior-independent data-based calibration remains an open problem.

17 pages, 7 figures, code available at https://github.com/hep-lbdl/calibrationpriors v2: Minor updates to match journal version

References in corpus (20)

• Array Programming with NumPy
• PYTHIA 6.4 Physics and Manual
• Particle-flow reconstruction and global event description with the CMS detector
• Jet energy scale and resolution in the CMS experiment in pp collisions at 8 TeV
• Performance of electron reconstruction and selection with the CMS detector in proton-proton collisions at sqrt(s) = 8 TeV
• Identification and energy calibration of hadronically decaying tau leptons with the ATLAS experiment in $pp$ collisions at $\sqrt{s}$=8 TeV
• Pileup Mitigation with Machine Learning (PUMML)
• Nuisance hardened data compression for fast likelihood-free inference
• Performance of the reconstruction and identification of high-momentum muons in proton-proton collisions at $\sqrt{s}=$ 13 TeV
• Uncertainty Aware Learning for High Energy Physics
• A Cautionary Tale of Decorrelating Theory Uncertainties
• On the Use of Neural Networks for Energy Reconstruction in High-granularity Calorimeters
• Pile-Up Mitigation using Attention
• Learning Uncertainties the Frequentist Way: Calibration and Correlation in High Energy Physics
• Transport away your problems: Calibrating stochastic simulations with optimal transport
• Calorimetric Measurement of Multi-TeV Muons via Deep Regression
• Convolutional Neural Networks for Shower Energy Prediction in Liquid Argon Time Projection Chambers
• Neural Conditional Reweighting
• Energy reconstruction in a liquid argon calorimeter cell using convolutional neural networks
• New approaches for boosting to uniformity

Cited by in corpus (1)

• Learning Uncertainties the Frequentist Way: Calibration and Correlation in High Energy Physics