Papers

Conservative neural posterior estimation via distributionally robust training

W Laplante, Y Hikida, C Dellaporta, FX Briol… - arXiv preprint arXiv …, 2026 - arxiv.org
Statistics paper stat.ML Suggest

… Simulation-based inference with neural posterior estimation (NPE) often yields overconfident and unreliable posteriors under limited simulation budgets. To address this, …

Link to paper

BibTeX

@article{2605.28516v1,
Author        = {William Laplante and Yuga Hikida and Charita Dellaporta and François-Xavier Briol and Ayush Bharti},
Title         = {Conservative neural posterior estimation via distributionally robust training},
Eprint        = {2605.28516v1},
ArchivePrefix = {arXiv},
PrimaryClass  = {stat.ML},
Abstract      = {Simulation-based inference with neural posterior estimation (NPE) often yields overconfident and unreliable posteriors under limited simulation budgets. To address this, we propose DRO-NPE, a distributionally robust approach that replaces the standard NPE objective with a worst-case loss over a Wasserstein ambiguity set. We introduce KL-based metrics for miscoverage and miscalibration, and use these to show that the DRO-NPE objective controls overfitting and reduces posterior overconfidence. Our method is tractable, parallelisable, and readily integrates with standard normalising flows. Across benchmark SBI tasks, DRO-NPE consistently improves coverage and calibration, while narrowing the gap between empirical and population NPE loss, leading to more reliable inference in low-simulation regimes.},
Year          = {2026},
Month         = {May},
Url           = {http://arxiv.org/abs/2605.28516v1},
File          = {2605.28516v1.pdf}
}

Share