Conservative neural posterior estimation via distributionally robust training
For practitioners of simulation-based inference with limited simulation budgets, DRO-NPE provides more reliable posterior estimates by controlling overconfidence and improving calibration.
DRO-NPE replaces the standard neural posterior estimation objective with a distributionally robust loss over a Wasserstein ambiguity set, improving coverage and calibration in low-simulation regimes. It consistently reduces overconfidence and overfitting across benchmark SBI tasks.
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.