PEtabTraining.jl
May 12, 2026 ยท View on GitHub
Training strategies for parameter estimation in dynamic models
PEtabTraining.jl implements training strategies that improve the efficiency of parameter
estimation for both ordinary differential equation (ODE) models and scientific machine
learning (SciML) models. The package is designed to be used with
PEtab.jl: training strategies can be applied
directly to a PEtabODEProblem to obtain modified training objectives (e.g. multiple
shooting objective) with a single line of code.
Currently, two training strategies are implemented:
- Curriculum learning: strategy where problem difficulty is progressively increased across curriculum stages. For dynamic models, this is typically done by gradually increasing the number of measurement time points (and often the simulation end time) over a fixed number of stages.
- Multiple shooting: strategy where the ODE simulation time span is split into windows that are fitted jointly. Each window has its own estimated initial state, and a continuity penalty is used to promote continuity between adjacent windows.
Concrete examples of how to apply these training strategies are available in the PEtab.jl documentation, while the PEtabTraining.jl documentation contains the detailed API and options for each strategy.
Installation
PEtabTraining.jl is a registered Julia package and can be installed via the Julia package manager:
import Pkg
Pkg.add("PEtabTraining")
PEtabTraining.jl is compatible with Julia 1.10 and above. For additional installation details, see the online documentation.
Citation
If you use PEtabTraining.jl in work that is published, please cite the paper below:
@article{PEtabBioinformatics2025,
title={PEtab.jl: advancing the efficiency and utility of dynamic modelling},
author={Persson, Sebastian and Fr{\"o}hlich, Fabian and Grein, Stephan and Loman, Torkel and Ognissanti, Damiano and Hasselgren, Viktor and Hasenauer, Jan and Cvijovic, Marija},
journal={Bioinformatics},
volume={41},
number={9},
pages={btaf497},
year={2025},
publisher={Oxford University Press}
}