Get Started
Home
Topics
Search
Library
Research questionHow can we estimate simulator-parameter distributions that reproduce observations without inaccurate fixed likelihood surrogates?A fixed likelihood surrogate can be inaccurate in parameter regions not covered by its training proposal. Fitting the source distribution to that surrogate may therefore reproduce surrogate errors rather than the simulator’s behavior.
AI
Machine Learning
Research Paper
Statistical Machine Learning
Latest papersRecent research connected to this question, newest first.Source Distribution Estimation by Posterior AveragingThe source evaluates expectation-maximization-based posterior averaging with separate source and posterior flows or a shared conditional flow. Evidence comes from three benchmark tasks under broad and misspecified initial priors, with comparisons against fixed-surrogate and iterated baselines; behavior beyond these settings is not established.research paper · Sep 2, 2026
Related questions
How can storm-surge emulators generate diverse spatial scenarios for hypothetical storms without costly physics simulations?How can compact device surrogates stay physically consistent and accurate across unseen process or geometry splits with scarce data?How can parametric PDE solution operators extrapolate to unseen regimes without failing silently?How can aerodynamic surrogates trained on CFD be grounded in scarce wind-tunnel measurements without retraining?