Provenance Graph Integration in Scientific Machine Learning Multi-workflows

• Débora Pina Universidade Federal do Rio de Janeiro (UFRJ)
• Lucas Moreira Dias Universidade Federal do Rio de Janeiro (UFRJ)
• Maria Célia Santos Lopes Universidade Federal do Rio de Janeiro (UFRJ)
• Daniel de Oliveira Universidade Federal Fluminense (UFF)
• Marta Mattoso Universidade Federal do Rio de Janeiro (UFRJ)

DOI: https://doi.org/10.5753/sbbd.2026.249458

Resumo

Scientific Machine Learning (SciML) multi-workflows consist of multiple independent workflows, often implemented as fragmented Jupyter Notebooks, which hinder provenance tracking and limit end-to-end querying and reproducibility. In this paper, we present a real-world SciML multi-workflow that advances methods for capturing and integrating provenance in SciML applications. The proposed approach combines scientific domain modeling of multi-workflows as prospective provenance, provenance capture mechanisms, and shared execution identifiers to enable the construction of an end-to-end provenance graph. We demonstrate the approach through queries over the integrated provenance graph of a multi-workflow seismic binary data segmentation application, highlighting its potential for long-term provenance management.

Palavras-chave: scientific machine learning, multi-workflow, provenance, W3C PROV

Referências

AlSalmi, H. and Elsheikh, A. H. (2023). Automated seismic semantic segmentation using attention u-net. Geophysics, 89(1):WA247–WA263.

Ferreira da Silva, R. et al. (2024). Workflows community summit 2024: Future trends and challenges in scientific workflows. (ORNL/TM-2024/3573).

Freire, J., Koop, D., Santos, E., and Silva, C. T. (2008). Provenance for Computational Tasks: A Survey. Computing in Science and Engineering, 10:11–21.

Gregori, L. et al. (2025). An llm-guided platform for multi-granular collection and management of data provenance. Journal of Big Data, 12(1):187.

Islam, M. S. U. and Wali, A. (2024). A comprehensive review of deep learning techniques for salt dome segmentation in seismic images. Journal of Applied Geophysics, 230:105504.

Liu, H., Wang, Y., Fan, W., Liu, X., Li, Y., Jain, S., Liu, Y., Jain, A., and Tang, J. (2022). Trustworthy ai: A computational perspective. ACM Trans. Intell. Syst. Technol., 14(1).

Missier, P. et al. (2025). From XAI to XEE: explainable end-to-end using influence and provenance. In Symposium on Advanced Database Systems, pages 560–569.

Moreau, L., Groth, P., Cheney, J., Lebo, T., and Miles, S. (2015). The rationale of PROV. J. Web Semant., 35:235–257.

Pina, D., Chapman, A., Kunstmann, L., de Oliveira, D., and Mattoso, M. (2024). Dlprov: A data-centric support for deep learning workflow analyses. In Proceedings of the Eighth Workshop on Data Management for End-to-End Machine Learning, page 77–85. ACM.

Pina, D., Kunstmann, L., et al. (2025). Dlprov: a suite of provenance services for deep learning workflow analyses. PeerJ Comp. Sci., 11:e2985.

Procko, T., Vonder Haar, L., and Ochoa, O. (2025). A survey of machine learning lifecycle provenance: Models, approaches and tools.

Schlegel, M. and Sattler, K.-U. (2025). Capturing end-to-end provenance for machine learning pipelines. Information Systems, 132:102495.

Souza, R. et al. (2022). Workflow provenance in the lifecycle of scientific machine learning. Concurrency and Computation: Practice and Experience, 34(14):e6544.

Souza, R. et al. (2023). Towards lightweight data integration using multi-workflow provenance and data observability. In International Conference on e-Science, pages 1–10.

Souza, R. et al. (2024). Workflow provenance in the computing continuum for responsible, trustworthy, and energy-efficient ai. In International Conference on e-Science, pages 1–7.

Souza, R. et al. (2025). Prov-agent: Unified provenance for tracking ai agent interactions in agentic workflows. In International Conference on eScience, pages 467–473.

Woyames, P. et al. (2025). Avaliação da capacidade de llms para especificar workflows. In Anais do XIX Brazilian e-Science Workshop, pages 81–88.