ISSUES IN BAYESIAN-ANALYSIS OF NEURAL-NETWORK MODELS

Authors
MULLER P INSUA DR
Citation
P. Muller et Dr. Insua, ISSUES IN BAYESIAN-ANALYSIS OF NEURAL-NETWORK MODELS, Neural computation, 10(3), 1998, pp. 749-770
Citations number
31
Categorie Soggetti
Computer Science Artificial Intelligence","Computer Science Artificial Intelligence
Journal title
Neural computation → ACNP
ISSN journal
08997667
Volume
10
Issue
3
Year of publication
1998
Pages
749 - 770
Database
ISI
SICI code
0899-7667(1998)10:3<749:IIBONM>2.0.ZU;2-L
Abstract
Stemming from work by Buntine and Weigend (1991) and MacKay (1992), th ere is a growing interest in Bayesian analysis of neural network model s. Although conceptually simple, this problem is computationally invol ved. We suggest a very efficient Markov chain Monte Carlo scheme for i nference and prediction with fixed-architecture feedforward neural net works. The scheme is then extended to the variable architecture case, providing a data-driven procedure to identify sensible architectures.