Figure 1
Performance of different machine learning models evaluated using the \(Z_0\) metric in a fivefold cross-validation. In (a), the models are trained on different signal and background ratios and tested on an equal number of signal and background events. An overview of the generated data set can be found in Table 2 in the experimental setup. In (b), the model complexity is increased while using the balanced data set for training and testing. We tested a convolutional neural network (CNN), a multilayer perceptron (MLP), the DirectRanker, and its combination with convolutional layers (CNN + DirectRanker).
