JADAI: Jointly Amortizing Adaptive Design and Bayesian Inference
Niels Bracher, Lars Kühmichel, Desi Ivanova, Xavier Intes, Paul Buerkner, Stefan Radev
Abstract
We consider problems of parameter estimation where design variables can be actively optimized to maximize information gain. To this end, we introduce JADAI, a framework that jointly amortizes Bayesian adaptive design and inference by training a policy, a history network, and an inference network end-to-end. The networks minimize a generic loss that aggregates incremental reductions in posterior error along experimental sequences without density evaluations. Inference networks are instantiated with diffusion models that can approximate high-dimensional and multimodal posteriors at every experimental step. JADAI achieves superior or competitive performance across adaptive design benchmarks.
BibTeX
@inproceedings{
bracher2026jadai,
title={{JADAI}: Jointly Amortizing Adaptive Design and Bayesian Inference},
author={Niels Leif Bracher and Lars K{\"u}hmichel and Desi R. Ivanova and Xavier Intes and Paul-Christian B{\"u}rkner and Stefan T. Radev},
booktitle={Forty-third International Conference on Machine Learning},
year={2026},
url={https://openreview.net/forum?id=tUpQVC7LvO}
}