Neural network gradient Hamiltonian Monte Carlo.
Comput Stat
; 34(1): 281-299, 2019 Mar.
Article
en En
| MEDLINE
| ID: mdl-31695242
ABSTRACT
Hamiltonian Monte Carlo is a widely used algorithm for sampling from posterior distributions of complex Bayesian models. It can efficiently explore high-dimensional parameter spaces guided by simulated Hamiltonian flows. However, the algorithm requires repeated gradient calculations, and these computations become increasingly burdensome as data sets scale. We present a method to substantially reduce the computation burden by using a neural network to approximate the gradient. First, we prove that the proposed method still maintains convergence to the true distribution though the approximated gradient no longer comes from a Hamiltonian system. Second, we conduct experiments on synthetic examples and real data to validate the proposed method.
Texto completo:
1
Colección:
01-internacional
Base de datos:
MEDLINE
Contexto en salud:
1_ASSA2030
Problema de salud:
1_financiamento_saude
Idioma:
En
Revista:
Comput Stat
Año:
2019
Tipo del documento:
Article
País de afiliación:
Estados Unidos