Simon Mak

Monotonicity constraints paradoxically degrade Bayesian surrogate models in high dimensions, but pairing transformed additivity with sparse basis priors restores a dimension-free $O(n^{-1/3})$ posterior contraction rate with linear-time inference.