This dataset contains all relevant data to reproduce the results of the paper titled "Capturing reduced-order quantum many-body dynamics out of equilibrium via neural ordinary differential equations". It contains the exact time series data of the two-particle reduced density matrix (2RDM) for each parameter configuration of the parameter scan, with some additional quantities discussed in the paper as well. Additionally, the test metrics of the neural ODE models for the whole parameter scan as well as the hyper-parameter optimized constraint models are provided, which contain the predicted and target trajectory of the reduced representation of the 2RDM as well as some additional metrics.

The exact time series of the Fermi-Hubbard model contains six files per parameter configuration:

'cumu_blocks.dat' contains the absolute values of the spin up-up and up-down block of the two-particle cumulant as well as the absolute value of the kernel component and the whole three-particle cumulant of the spin up-up-down block over time 'energy_mcscf.dat' contains the correlation energy and a column called energy, where the first value of this column is required to calculate the initial potential energy 'energyspectrum_groundstate.dat' where the first value is used to calculate the initial potential energy 'Hubbard_all.den2ab' contains the spin up-down block of the 2RDM over time 'occ_numbers.doublon2' contains the doublon occupation numbers for each site over time 'occ_numbers.rho1' contains the occupation numbers for each site over time

The neural ODE test metrics for each parameter configuration are provided as well. Each file contains a short and long prediction of the normalized and reduced representation of the 2RDM data as well as the corresponding target values. Additionally, the Pearson correlation and some other metrics are given for the short prediction of 1000 time steps.

Similar test metrics files are provided for the hyper-parameter optimized constraint and non-constraint models as well as for the line scan.