Training Data for Gollum in Higgs Uncertainty Challenge

This is a subset of the training data for "Gollum", Team HEPHY's submission to the FAIR Universe Higgs Uncertainty Challenge. For the full training, we used systematic variations up to 3 standard deviations. All of these can be generated from the nominal file (provided by the challenge) and the code that the challenge provided. Data was taken from CodeBench: https://www.codabench.org/datasets/download/b9e59d0a-4db3-4da4-b1f8-3f609d1835b2/, and systematic variations were applied according to the description https://arxiv.org/abs/2410.02867. Gollum Code: https://github.com/HephyAnalysisSW/GOLLUM Gollum Publication: TBA FAIR Univerese Challenge: https://www.codabench.org/competitions/2977, https://github.com/FAIR-Universe/HEP-Challenge/tree/master/ The files are: nominal.h5 is a dataset without systematic variations met, jes, tes in the file name indicates a variation of MET (missing energy), jet energy scale, or tau energy scale to the value that follows. For example: the file tes_0p99_jes_0p99.h5 includes events where the tau and jet energy scales were both multiplied with a factor of 0.99, corresponding to nuicances parameters with the value of -1. Normalization-type uncertainties are not included, as these samples can be obtained by changing the corresponding event weights. The file toy_mu_2.h5 is a toy dataset with true value of the signal strength mu set to 2. The detailed values of the nuisance parameters (in the format (1+alpha), see the documentation of the Challenge) are stored in a dictionary that can be found in toy_mu_2.pkl: import pickle myfile = open('toy_mu_2.pkl', 'rb')nuisances = pickle.load(myfile)print(nuisances)>>> {'mu': 2.0, 'tes': 0.9945927027846702, 'jes': 1.0094985245559578, 'soft_met': 0.16120480304203255, 'ttbar_scale': 0.9866736788084463, 'diboson_scale': 1.0610956857813547, 'bkg_scale': 0.999100305445466}