talks

EuroSciPy 2025 talk (jointly with Seth Axen) bridging sbi and Pyro to enable simulation-based hierarchical Bayesian inference. SBI handles the intractable-likelihood simulator; Pyro handles hierarchical structure. Demonstrated on a “cookie factory” worked example and on hierarchical drift-diffusion models for decision-making.

• Slides + code: janfb/pyro-meets-sbi
• Conference page: EuroSciPy 2025 talk
• Cookie-factory inspiration: Juan Camilo Orduz’s blog post

EuroSciPy 2024 talk introducing simulation-based inference and the sbi library to the scientific Python community: state-of-the-art SBI algorithms, diagnostics for accuracy checks, and applications across neuroscience and astrophysics.

• Conference page: EuroSciPy 2024 talk

Internal seminar discussing the gap between SBI methodological advances and real-world deployment: an overview of NPE/NLE/NRE-style methods, why broader adoption is still hard, how the sbi package addresses usability, and case studies in neuroscience, epidemiology, and climate.

Invited talk at the BCCN Berlin seminar series on applying simulation-based inference to neuroscience. Covered Mixed Neural Likelihood Estimation (MNLE) for decision-making models, SBI for connectomics, and tooling via the sbi package.