This repository contains the code for running and reproducing the experiments from the paper JANA: Jointly Amortized Neural Approximation of Complex Bayesian Models, presented at the Conference on Uncertainty in Artificial Intelligence (UAI 2023).
JANA lets you train and validate specialized neural networks for simultaneously amortized simulation-based inference and surrogate modeling in a Bayesian framework. The method is described in our paper:
Radev, S. T., Schmitt, M., Pratz, V., Picchini, U., Köthe, U., & Bürkner, P. C. (2023). JANA: Jointly Amortized Neural Approximation of Complex Bayesian Models. arXiv preprint arXiv:2302.09125, available for free at: https://arxiv.org/abs/2302.09125.
The code depends on the BayesFlow library, which implements all JANA components and benchmark simulators. Each experiment features self-contained code and individual instructions. Checkpoints for most networks are provided at the cost of the repository's size.
You can easily cite the proceedings paper as:
@InProceedings{radev2023jana,
title = {{JANA}: Jointly amortized neural approximation of complex {B}ayesian models},
author = {Radev, Stefan T. and Schmitt, Marvin and Pratz, Valentin and Picchini, Umberto and K\"othe, Ullrich and B\"urkner, Paul-Christian},
booktitle = {Proceedings of the 39th Conference on Uncertainty in Artificial Intelligence},
pages = {1695--1706},
year = {2023},
volume = {216},
series = {Proceedings of Machine Learning Research},
publisher = {PMLR},
}This work was supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy -– EXC-2181 - 390900948 (the Heidelberg Cluster of Excellence STRUCTURES) and -- EXC-2075 - 390740016 (the Stuttgart Cluster of Excellence SimTech), the Cyber Valley Research Fund (grant number: CyVy-RF-2021-16), the Swedish National Research Council (Vetenskapsrådet 2019-03924), the Chalmers AI Research Centre, the Informatics for Life initiative funded by the Klaus Tschira Foundation and Google Cloud through the Academic Research Grants program.
MIT