Models such as feed-forward neural networks and certain other structures investigated in the computer science literature are not amenable to closed-form Bayesian analysis. The paper reviews the various approaches taken to overcome this difficulty, involving the use of Gaussian approximations, Markov chain Monte Carlo simulation routines and a class of non-Gaussian but “deterministic” approximations called variational approximations.
Publié le : 2004-02-14
Classification:
Bayesian methods,
Bayesian model choice,
feed-forward neural network,
graphical model,
Laplace approximation,
machine learning,
Markov chain Monte Carlo,
variational approximation
@article{1089808278,
author = {Titterington, D. M.},
title = {Bayesian Methods for Neural Networks and Related Models},
journal = {Statist. Sci.},
volume = {19},
number = {1},
year = {2004},
pages = { 128-139},
language = {en},
url = {http://dml.mathdoc.fr/item/1089808278}
}
Titterington, D. M. Bayesian Methods for Neural Networks and Related Models. Statist. Sci., Tome 19 (2004) no. 1, pp. 128-139. http://gdmltest.u-ga.fr/item/1089808278/