Addressing Misspecification in Simulation-based Inference through Data-driven Calibration

Antoine Wehenkel, Juan L. Gamella, Ozan Sener, Jens Behrmann, Guillermo Sapiro, Jörn Jacobsen, Marco Cuturi

International Conference on Machine Learning 2025 · Oral

This talk by Juan L. Gamella and Antoine Wehenkel, representing a collaborative effort from Apple, introduces **Warped Posterior Estimator (WORP)**, a novel framework designed to enhance **Simulation-Based Inference (SBI)** in scenarios where simulators are inherently misspecified. The core challenge in many scientific and engineering domains is the need to infer complex, often unobservable parameters from readily available observations, a task complicated by a severe scarcity of labeled data. While simulators offer a powerful avenue for incorporating domain knowledge and generating synthetic data, their inherent inaccuracies—the "sim-to-real gap"—typically render standard SBI pipelines unreliable, leading to inaccurate and uncalibrated uncertainty estimates.

AI review

WORP is a competent and well-motivated contribution to simulation-based inference under misspecification. The core idea — use optimal transport in embedding space to prevent the neural density estimator from operating out-of-distribution — is clean and practically relevant. The 20x data efficiency result on the causal chamber is the strongest empirical signal in the paper. However, the theoretical guarantees appear thin: the key modeling assumption (misspecification depends on θ only through X_s) is load-bearing but its implications for posterior consistency are not formally analyzed. The…