Parameter uncertainties for imperfect surrogate models in the low-noise regime

Thomas D Swinburne, Danny Perez

Bayesian regression determines model parameters by minimizing the expected loss, an upper bound to the true generalization error. However, the loss ignores misspecification, where…