Bayesian methods for neural networks and related models

被引：107

作者：

Titterington, DM ^{[1
]}

机构：

[1] Univ Glasgow, Dept Stat, Glasgow G12 8QQ, Lanark, Scotland

来源：

STATISTICAL SCIENCE | 2004年 / 19卷 / 01期

关键词：

Bayesian methods; Bayesian model choice; feed-forward neural network; graphical model; Laplace approximation; machine learning; Markov chain Monte Carlo; variational approximation;

D O I：

10.1214/088342304000000099

中图分类号：

O21 [概率论与数理统计]; C8 [统计学];

学科分类号：

020208 ; 070103 ; 0714 ;

摘要：

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.

引用

页码：128 / 139

页数：12

共 98 条

[1] Robust full Bayesian learning for radial basis networks [J].

Andrieu, C ;

de Freitas, N ;

Doucet, A .

NEURAL COMPUTATION, 2001, 13 (10) :2359-2407

[2]

Andrieu C, 2000, ADV NEUR IN, V12, P379

[3]

[Anonymous], P 17 C UNC ART INT

[4] Parameter estimation for hidden Markov chains [J].

Archer, GEB ;

Titterington, DM .

JOURNAL OF STATISTICAL PLANNING AND INFERENCE, 2002, 108 (1-2) :365-390

[5] Independent factor analysis [J].

Attias, H .

NEURAL COMPUTATION, 1999, 11 (04) :803-851

[6]

Attias H, 1999, UNCERTAINTY IN ARTIFICIAL INTELLIGENCE, PROCEEDINGS, P21

[7]

Attias H, 2000, ADV NEUR IN, V12, P209

[8]

Barber D., 1998, Neural Networks and Machine Learning. Proceedings, P215

[9]

BERGER J. O., 2013, Statistical Decision Theory and Bayesian Analysis, DOI [10.1007/978-1-4757-4286-2, DOI 10.1007/978-1-4757-4286-2]

[10]

Bishop C. M., 2000, P 16 C UNC ART INT S, P46

← 1 2 3 4 5 6 7 8 9 10 →