No abstract

In the past few years, there has been considerable activity in both academic and industrial research to develop innovative machine learning approaches to locate novel, high-performing molecules in chemical space....

Physical theories that depend on many parameters or are tested against data from many different experiments pose unique challenges to statistical inference. Many models in particle physics, astrophysics and cosmology fall into one or both of these categories. These issues are often sidestepped with statistically unsound ad hoc methods, involving intersection of parameter intervals estimated by multiple experiments, and random or grid sampling of model parameters. Whilst these methods are easy to apply, they exhibit pathologies even in low-dimensional parameter spaces, and quickly become problematic to use and interpret in higher dimensions. In this article we give clear guidance for going beyond these procedures, suggesting where possible simple methods for performing statistically sound inference, and recommendations of readily-available software tools and standards that can assist in doing so. Our aim is to provide any physicists lacking comprehensive statistical training with recommendations for reaching correct scientific conclusions, with only a modest increase in analysis burden. Our examples can be reproduced with the code publicly available at Zenodo.

The evaluation of higher-order cross-sections is an important component in the search for new physics, both at hadron colliders and elsewhere. For most new physics processes of interest, total cross-sections are known at next-to-leading order (NLO) in the strong coupling $$\alpha _s$$ α s , and often beyond, via either higher-order terms at fixed powers of $$\alpha _s$$ α s , or multi-emission resummation. However, the computation time for such higher-order cross-sections is prohibitively expensive, and precludes efficient evaluation in parameter-space scans beyond two dimensions. Here we describe the software tool , which allows for fast evaluation of cross-sections based on the use of machine-learning regression, using distributed Gaussian processes trained on a pre-generated sample of parameter points. This first version of the code provides all NLO Minimal Supersymmetric Standard Model strong-production cross-sections at the LHC, for individual flavour final states, evaluated in a fraction of a second. Moreover, it calculates regression errors, as well as estimates of errors from higher-order contributions, from uncertainties in the parton distribution functions, and from the value of $$\alpha _s$$ α s . While we focus on a specific phenomenological model of supersymmetry, the method readily generalises to any process where it is possible to generate a sufficient training sample.