This repository provides a collection of real-data scientific machine learning (SciML) benchmark problems in the PEtab-SciML data format. The models combine mechanistic ordinary differential equation (ODE) and machine learning models and are intended to support the development and evaluation of methodology for SciML problems.
Contributions of new benchmark problems are welcome. See CONTRIBUTING.md for details.
Benchmark models are available in Benchmark-Models. In addition to the
PEtab-SciML problem files, each model directory contains an expected.yaml file with
expected loss or likelihood values for the provided parameter values.
Current models include:
| Model | Hybridization | Est. params | Mech. params | ML params | ODE states |
|---|---|---|---|---|---|
| Dandekar_Patterns2020 | UDE | 54 | 3 | 51 | 4 |
| Ko_NeurIPS2023 | Neural ODE | 2902 | 0 | 2902 | 4 |
Clone the repository from GitHub:
git clone https://github.com/sebapersson/Benchmark-Models-PEtab-SciML.gitAll files required for testing are included in the repository.
If you encounter problems, please open an issue on GitHub.