Speaker
Description
Reinterpreting the LHC results as bounds on the Wilson Coefficients (WCs) of the Standard Model Effective Field Theory (SMEFT) allows studying new-physics effects in a model-independent way with minimal assumptions. However, the large number of effective interactions along with theoretical and experimental uncertainties result in poor constraints on WCs that motivate the use of alternative techniques with more comprehensive extraction of information from data. In this presentation, I will talk about constructing physics-inspired graphs from the final states of $p p \rightarrow t \bar{t}$ production with semi-leptonic top decays, and using Edge Convolution Neural Networks in order to condense the multidimensional phase space information. When using the output of the Neural Network to identify a signal region such that the SM contribution is minimised, the approach yields improvements on the bounds of WCs, compared to analyses on inclusive collision data employing differential distributions to measure deviations from the SM.
