Datasets for Self-Driving Trigger Study at L1

These datasets are generated from CMS 2016 open datasets for L1 hadronic objects (jets). Specific process files (MinBias, HToAATo4B, etc) include reconstructed jets information, and the number of primary vertices (NPV) in each event, which are our starting point samples. Some of these files are used to train and test an Anomaly Detection model, and some of them are our benchmark study inputs for control algorithms (which our Anomaly Detection models have never seen). Trigger_food files include anomaly score, HT (Hadronic Transverse Momentum), number of jets and NPV for each event of each process (we have separated files for simulated background vs real data), for control algorithms studies. 

MinBias_1.h5: This dataset is used for Anomaly Detection training in studies with a simulated sample as background. 

MinBias_2.h5: simulated background sample used for control studies when focusing on only MC simulated cases.

TT_1.h5: simulated signal sample for hadronic decay of the Standard Model ttbar process

HToAATo4B.h5: simulated signal sample for Beyond Standard Model hadronic decay of Higgs particle to unknown particle of A, which decays to B jets.

data_Run_2016_283408_longest.h5: longest run in the sample, used for studying the control algorithm when real data is the background.

data_Run_2016_283876.h5: second-longest run used for training the Anomaly Detection model, where real data is involved. 

Trigger_food_MC.h5: Control-variables dataset, including only necessary information for running trigger code, such as anomaly scores, HT, NPV, etc, for different simulated processes. Using this dataset increases the speed of the code, as anomaly scores are generated once. 

Trigger_food_Data.h5: Control-variables dataset, with real data as the background and matched MC simulated signal and real background based on NPV.