MLPF focuses on developing full event reconstruction based on computationally scalable and flexible end-to-end ML models.
The following table specifies which version of the jpata/particleflow software should be used with which version of the tensorflow datasets.
| Code | CMS dataset | CLIC dataset |
|---|---|---|
| 1.9.0 | 2.4.0 | 2.2.0 |
| 2.0.0 | 2.4.0 | 2.3.0 |
| 2.1.0 | 2.5.0 | 2.5.0 |
| 2.2.0 | 2.5.0 | 2.5.0 |
- paper: https://doi.org/10.1038/s42005-024-01599-5
- code: https://doi.org/10.5281/zenodo.10893930
- dataset: https://doi.org/10.5281/zenodo.8409592
- results: https://doi.org/10.5281/zenodo.10567397
- weights: https://huggingface.co/jpata/particleflow/tree/main/clic/clusters/v1.6
The following datasets are available to reproduce the studies. They include full Geant4 simulation and reconstruction based on the CLIC detector. We have no affiliation with the CLIC collaboration, therefore these datasets are to be used only for computational studies and come with no warranty.
- MLPF-CLIC, raw data: https://zenodo.org/records/8260741 or https://www.coe-raise.eu/od-pfr
- MLPF-CLIC, processed for ML, tracks and clusters: https://zenodo.org/records/8409592
- MLPF-CLIC, processed for ML, tracks and hits: https://zenodo.org/records/8414225
- ACAT 2022:
- CERN-CMS-DP-2022-061, http://cds.cern.ch/record/2842375
- ACAT 2021:
- J. Phys. Conf. Ser. 2438 012100, http://dx.doi.org/10.1088/1742-6596/2438/1/012100
- CERN-CMS-DP-2021-030, https://cds.cern.ch/record/2792320
- paper: https://doi.org/10.1140/epjc/s10052-021-09158-w
- code: https://doi.org/10.5281/zenodo.4559587
- dataset: https://doi.org/10.5281/zenodo.4559324
You are welcome to reuse the code in your work in accordance with the license.
For academic work, please consider citing the following papers:
- initial idea with scalable GNN, code v1.1: https://doi.org/10.1140/epjc/s10052-021-09158-w
- improved event-level performance in full simulation, code v1.6.2: https://doi.org/10.1038/s42005-024-01599-5
- studies in CMS: https://cds.cern.ch/record/2792320, http://dx.doi.org/10.1088/1742-6596/2438/1/012100, http://cds.cern.ch/record/2842375
If you use the code in a significant way for research purposes, please consider citing the tagged version that you used, for example:
- Joosep Pata, Eric Wulff, Farouk Mokhtar, Javier Duarte, Aadi Tepper, Ka Wa Ho, & Lars Sørlie. (2025). jpata/particleflow: v2.2.0 (v2.2.0). Zenodo. https://doi.org/10.5281/zenodo.14650991
If you use the datasets prepared by the MLPF team for academic work, please cite the appropriate dataset via the zenodo link, as well as the corresponding paper.
At the moment, we are unable to release work-in-progress datasets before the corresponding academic publication is out. If you have a collaboration idea that does not fit into the above categories, please get in touch!