Physical dataset for OneFlow data/MC corrections

This dataset contains two files (data and MC) with one million entries each. The synthetic samples resemble typical distributions in high-energy physics and emulate a mismodelling of Monte-Carlo simulated distributions that do not fully agree with data. The samples have one index and seven features (pT, eta, noise, v1A, v2A, v1B, v2B). The dataset features non-trivial correlations and conditioning of the latter four v-features on the first three variables pT, eta, and noise. The v1A and v2A features are Gaussian-like, whereas the v1B and v2B features are discontinuous.

This dataset is well suited to explore correction methods, both morphing and reweighting, in a challenging setup.

This dataset was used in the publication One Flow to Correct Them all: Improving Simulations in High-Energy Physics with a Single Normalising Flow and a Switch to demonstrate the suitability of a single normalising flow to morph the MC distribution into the data distribution. For more details on this, please refer to the publication.

This dataset was generated using publicly available code at GitHub:OneFlowPublic