A0600
Title: Bayesian optimization of hyperparameters when the marginal likelihood is estimated by MCMC
Authors: Oskar Gustafsson - Stockholm University (Sweden)
Mattias Villani - Stockholm University (Sweden) [presenting]
Par Stockhammar - Stockholm University (Sweden)
Abstract: Bayesian models in econometrics often involve a small set of hyperparameters determined by maximizing the marginal likelihood. Bayesian optimization is a popular iterative method where a Gaussian process posterior of the underlying function is sequentially updated by new function evaluations. An acquisition strategy uses this posterior distribution to decide where to place the next function evaluation. We propose a novel Bayesian optimization framework for situations where the user controls the computational effort, and therefore the precision of the function evaluations. This is a common situation in econometrics where the marginal likelihood is often computed by Markov Chain Monte Carlo (MCMC) methods, with the precision determined by the number of MCMC draws. The method is used to find optimal prior hyperparameters in the steady-state vector autoregression fitted to US macroeconomic data.
