On MCMC sampling in Bayesian MLP neural networks

Author

Vehtari, Aki ; Särkkä, Simo ; Lampinen, Jouko

Author_Institution

Dept. of Comput. Eng., Helsinki Univ. of Technol., Espoo, Finland

Volume

1

fYear

2000

fDate

2000

Firstpage

317

Abstract

Bayesian MLP neural networks are a flexible tool in complex nonlinear problems. The approach is complicated by need to evaluate integrals over high-dimensional probability distributions. The integrals are generally approximated with Markov chain Monte Carlo (MCMC) methods. There are several practical issues which arise when implementing MCMC. This article discusses the choice of starting values and the number of chains in Bayesian MLP models. We propose a new method for choosing the starting values based on early stopping and we demonstrate the benefits of using several independent chains

Keywords

Bayes methods; Markov processes; Monte Carlo methods; multilayer perceptrons; Bayesian MLP neural networks; MCMC sampling; Markov chain Monte Carlo methods; complex nonlinear problems; high-dimensional probability distributions; independent chains; integrals; Bayesian methods; Intelligent networks; Laboratories; Minimization methods; Monte Carlo methods; Neural networks; Probability distribution; Sampling methods; Uncertainty;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks, 2000. IJCNN 2000, Proceedings of the IEEE-INNS-ENNS International Joint Conference on

Conference_Location

Como

ISSN

1098-7576

Print_ISBN

0-7695-0619-4

Type

conf

DOI

10.1109/IJCNN.2000.857855

Filename

857855