Data for Learning Missing Physics from Legacy Simulators with Alternating Neural Integrators – Part 3: Extended Experiments

This record is Part 3 of the data for the manuscript Learning missing physics from legacy simulators with alternating neural integrators. The package is prepared for manuscript review and reproducibility. The source code is available from the accompanying GitHub repository: https://github.com/shanxue-w/ANI Due to file size limitations and additional experiments introduced during revision, the complete reproducibility package is split across three Zenodo records. This Part 3 record contains supplementary datasets for additional experiments conducted during the peer-review process. While Part 1 and Part 2 contain the original materials for the FitzHugh-Nagumo, Kolmogorov flow, and initial Navier-Stokes experiments, this part focuses on revised and extended experimental results introduced during revision, together with two real-data experiments. Users should refer to all three Zenodo records for a complete reproduction of the manuscript results. Part 1 serves as the primary entry point to the complete package. Contents of Part 3 include: Revised Lorenz-Stenflo experiments. The results in this record should be regarded as the definitive version for the manuscript. Revised Kolmogorov flow experiments. The results in this record correspond to the final main-text results. Additional FitzHugh-Nagumo revision experiments used for supplementary results. Additional Navier-Stokes materials, with execution commands consistent with those used in Part 2. HPC experiment dataset. Battery degradation experiment based on cell B0005. Basin hydrology experiment based on the CAMELS dataset. Related records: Link to Part 1: https://doi.org/10.5281/zenodo.17412715 Link to Part 2: https://doi.org/10.5281/zenodo.17412698 License This repository contains the processed research data. The source code is available from the accompanying GitHub repository: https://github.com/shanxue-w/ANI Source code: MIT License. Processed datasets, figure/source data, tables, and reproducibility materials: Creative Commons Attribution 4.0 International License (CC BY 4.0), unless otherwise stated. Third-party source datasets retain their original licenses and terms of use.