Abstract
Machine learning techniques are becoming an integral component of data analysis in High Energy Physics (HEP). These tools provide a significant improvement in sensitivity over traditional analyses by exploiting subtle patterns in high-dimensional feature spaces. These subtle patterns may not be well-modeled by the simulations used for training machine learning methods, resulting in an enhanced sensitivity to systematic uncertainties. Contrary to the traditional wisdom of constructing an analysis strategy that is invariant to systematic uncertainties, we study the use of a classifier that is fully aware of uncertainties and their corresponding nuisance parameters. We show that this dependence can actually enhance the sensitivity to parameters of interest. Studies are performed using a synthetic Gaussian dataset as well as a more realistic HEP dataset based on Higgs boson decays to tau leptons. For both cases, we show that the uncertainty aware approach can achieve a better sensitivity than alternative machine learning strategies.
Highlights
The usefulness of physical measurements is tied to the magnitude and reliability of their estimated uncertainties
To evaluate the power of each approach above, we apply them to a common use case, fitting a signal hypothesis in the presence of background, where both signal and background depend on nuisance parameters
In this paper,3 we have advocated for uncertainty-aware classifiers where the dependence on nuisance parameter is maximized during training by exploiting parametrized classifiers [29,30]
Summary
The usefulness of physical measurements is tied to the magnitude and reliability of their estimated uncertainties. The challenging task of quantifying the systematic uncertainty on a measured physical parameter has become even more important due to the growing use of complex statistical procedures based on modern machine learning [3,4,5,6,7,8,9]. We focus on only the construction of classifiers as useful statistics for downstream analysis and not on full likelihood (ratio) estimation In this way, our uncertainty-aware classifier approach is a straightforward extension of existing analyses performed at the Large Hadron Collider (LHC) and may result in immediate improvements in sensitivity. Our uncertainty-aware classifier approach is a straightforward extension of existing analyses performed at the Large Hadron Collider (LHC) and may result in immediate improvements in sensitivity This prescription allows for easy post hoc histogram-based diagnostics.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have