Skip to main content
Dryad

Data for: Foundations of a fast, data-driven, machine-learned simulator

Howard, Jessica N.1; Mandt, Stephan1; Whiteson, Daniel1; Yang, Yibo1

Published May 04, 2021 on Dryad. https://doi.org/10.7280/D1WQ3R

Data files

May 04, 2021 version files 241.03 MB

Abstract

We introduce Optimal-Transport-based Unfolding and Simulation (OTUS), a novel, fast simulator based on unsupervised machine-learning that is capable of predicting experimental data from theoretical models. Simulations are crucial in science because they map from theoretical models to experimental data, allowing scientists to test predictions of theoretical models against the reality of experiments.  Experimental data is often reconstructed from indirect measurements causing the aggregate transformation from theoretical models to experimental data to be poorly described by analytical methods. Scientists instead rely on ad-hoc, numerical simulations at great computational cost. Capable of learning directly from data, OTUS trains a probabilistic autoencoder to transform directly between theoretical models and experimental data. This is achieved by identifying the probabilistic autoencoder's latent space with the space of theoretical models, causing the decoder network to become a fast, predictive simulator with the potential to replace current, computationally costly simulators. Using particle physics as an illustrative example, we provide proof-of-principle results for Z-boson and top-quark decays, but stress that OTUS can be widely applied to other fields.