PEtabTraining.jl

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, three 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.

• Curriculum multiple shooting: strategy combining multiple shooting with a curriculum schedule. Training starts from a multiple-shooting formulation (often easier to optimize) and progressively reduces the number of windows until the original single-window problem is recovered.

Getting started

PEtabTraining.jl is meant to be used together with PEtab.jl, and all training strategies operate on a PEtabODEProblem. Concrete examples of how to apply the training strategies are available in the PEtab.jl documentation. A detailed overview of available options for each training method is provided in the API documentation.

Installation

PEtabTraining.jl is an officially registered Julia package, tested and supported on Linux, macOS and Windows. The easiest way to install it is via the Julia package manager. In the Julia REPL, enter:

julia> ] add PEtabTraining

or alternatively

julia> using Pkg; Pkg.add("PEtabTraining")

PEtabTraining is compatible with Julia 1.10 and above. For best performance, we strongly recommend using the latest Julia version, which can be most reliably installed using juliaup.

Getting help

If you have any problems using PEtabTraining, here are some helpful tips:

• Post your questions in the #sciml-sysbio channel on the Julia Slack.

• If you encounter unexpected behavior or a bug, please see how to file an issue on the Contributing page.

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}
}