Punzi-loss: A non-differentiable metric approximation for sensitivity optimisation in the search for new particles

Abstract

We present the novel implementation of a non-differentiable metric approximation and a corresponding loss-scheduling aimed at the search for new particles of unknown mass in high energy physics experiments. We call the loss-scheduling, based on the minimisation of a figure-of-merit related function typical of particle physics, a Punzi-loss function, and the neural network that utilises this loss function a Punzi-net. We show that the Punzi-net outperforms standard multivariate analysis techniques and generalises well to mass hypotheses for which it was not trained. This is achieved by training a single classifier that provides a coherent and optimal classification of all signal hypotheses over the whole search space. Our result constitutes a complementary approach to fully differentiable analyses in particle physics. We implemented this work using PyTorch and provide users full access to a public repository containing all the codes and a training example.

arXiv commentsubmitted to EPJC

References in corpus (6)

• The automated computation of tree-level and next-to-leading order differential cross sections, and their matching to parton shower simulations
• The Belle II Physics Book
• The Belle II Core Software
• INFERNO: Inference-Aware Neural Optimisation
• Search for an invisibly decaying $Z^{\prime }$ boson at Belle II in $e^{+}{e}^{-}\to {\mu }^{+}{\mu }^{-}(e^{\pm }{\mu }^{\mp })$ plus missing energy final states
• Optimal statistical inference in the presence of systematic uncertainties using neural network optimization based on binned Poisson likelihoods with nuisance parameters

Cited by in corpus (5)

• Search for an invisible $Z^{\prime }$ in a final state with two muons and missing energy at Belle II
• First measurement of A = 4 (anti)hypernuclei at the LHC
• Interpretable deep learning models for the inference and classification of LHC data
• First measurements of the branching fractions for the decay modes ${\Xi }_c^0\to \Lambda \eta$ and ${\Xi }_c^0\to \Lambda {\eta }^{\prime }$ and search for the decay ${\Xi }_c^0\to \Lambda {\pi }^0$ using Belle and Belle II data
• Towards an anomaly detection pipeline for gravitational waves at the Einstein Telescope