Optimal Statistical Inference in the Presence of Systematic Uncertainties Using Neural Network Optimization Based on Binned Poisson Likelihoods with Nuisance Parameters

• DOI: 10.1007/s41781-020-00049-5
• Publication Date: 2021-01-13T13:02:32Z
• Authors: Stefan Wunsch; Simon Jörger; Roger Wolf; Günter Quast
• Keywords: Statistical Inference; Feature (linguistics)

Abstract Data analysis in science, e.g., high-energy particle physics, is often subject to an intractable likelihood if the observables and observations span a high-dimensional input space. Typically problem solved by reducing dimensionality using feature engineering histograms, whereby latter allows build Poisson statistics. However, presence of systematic uncertainties represented nuisance parameters likelihood, optimal reduction with minimal loss information about interest not known. This work presents novel strategy construct neural networks for differential formulation histograms so that full workflow can be optimized result statistical inference, variance parameter interest, as objective. We discuss how this approach results estimate close applicability technique demonstrated simple example based on pseudo-experiments more complex from physics.