Bayesian inference with an adaptive prior based on the intrinsic spectral structure of the data can achieve consistent predictions in overparameterized nonlinear regression models.
