Slice sampling covariance hyperparameters of latent Gaussian models - Statistics > ComputationReport as inadecuate




Slice sampling covariance hyperparameters of latent Gaussian models - Statistics > Computation - Download this document for free, or read online. Document in PDF available to download.

Abstract: The Gaussian process GP is a popular way to specify dependencies betweenrandom variables in a probabilistic model. In the Bayesian framework thecovariance structure can be specified using unknown hyperparameters.Integrating over these hyperparameters considers different possibleexplanations for the data when making predictions. This integration is oftenperformed using Markov chain Monte Carlo MCMC sampling. However, withnon-Gaussian observations standard hyperparameter sampling approaches requirecareful tuning and may converge slowly. In this paper we present a slicesampling approach that requires little tuning while mixing well in both strong-and weak-data regimes.



Author: Iain Murray, Ryan Prescott Adams

Source: https://arxiv.org/







Related documents