Dates: 21-23 January 2026
Location: IMAG building, Université Grenoble Alpes campus, Grenoble, France
Duration: 3 days (1.5 days tutorial + 1.5 days hackathon)
Target audience: researchers, students, and practitioners interested in parameter estimation, model calibration, or uncertainty quantification in physical, biological, or engineering simulations.
Registration: Free, but limited spots. Please register by 31 December 2025. The organizers will get back to you to confirm your registration.
Environment: conda or mamba setup, preconfigured via provided environment.yml
Code & materials: shared through this repository and GitHub Classroom
Outputs: participants are encouraged to share notebooks, results, and ideas for follow-up collaborations.
Location: IMAG building, Université Grenoble Alpes campus, Grenoble, France
Duration: 3 days (1.5 days tutorial + 1.5 days hackathon)
Target audience: researchers, students, and practitioners interested in parameter estimation, model calibration, or uncertainty quantification in physical, biological, or engineering simulations.
Registration: Free, but limited spots. Please register by 31 December 2025. The organizers will get back to you to confirm your registration.
Programme
Day 1–2 (Tutorial, 1.5 days)
Led by Jan Teusen. Topics include:
- Generative modeling and the SBI framework
- Neural posterior estimation (NPE), likelihood and ratio estimation
- Hands-on notebooks using the sbi Python package
- Applying SBI to toy and benchmark simulators
Day 2–3 (Hackathon, 1.5 days)
Participants form small teams around real or toy simulators (your own, or provided examples).
Possible directions:
- Adapting SBI to your research simulator
- Benchmarking different inference strategies
- Building reproducible pipelines for calibration or uncertainty quantification
- Sharing and discussing results at the closing session
We will take inspiration from successful open hackathons such as the IGE-Jaxathon 2025, emphasizing collaboration, openness, and applied research.
Practical Information
Prerequisites: basic Python and machine learning familiarityEnvironment: conda or mamba setup, preconfigured via provided environment.yml
Code & materials: shared through this repository and GitHub Classroom
Outputs: participants are encouraged to share notebooks, results, and ideas for follow-up collaborations.