Performance Tests of $tH$(bb) Signal and Background Separation Using a Binary Classifier Neural Network

The production of a Higgs boson in association with a single top quark is a strongly suppressed process in the Standard Model (SM). In the current data set of 140\,fb$^{-1}$, the SM expected production rate is below the experimental sensitivity. Thus, observing such a $tH$ production would indicate new physics. The absolute $t\Bar{t}H$ coupling strength was already measured and the $tH$ process can in addition measure the relative sign of the $t\Bar{t}H$ coupling. Therefore, observing the $tH$ process gives an important additional insight into the physics of the Higgs mechanism. Owing to the low production rate, it is particularly important to enhance the signal sensitivity, and a Neural Network (NN) is used. The resulting significance is studied by varying the NN structure. Based on simulated data, the performances of these different NN structures were tested and results are expressed as area under the ROC curve to quantify the signal and background separation.