Nonparametric learning from Bayesian models with randomized objective functions

no code implementations • NeurIPS 2018 • Simon Lyddon, Stephen Walker, Chris C. Holmes

Bayesian learning is built on an assumption that the model space contains a true reflection of the data generating mechanism.

General Bayesian Updating and the Loss-Likelihood Bootstrap

no code implementations • 22 Sep 2017 • Simon Lyddon, Chris Holmes, Stephen Walker

In this paper we revisit the weighted likelihood bootstrap, a method that generates samples from an approximate Bayesian posterior of a parametric model.