NUTS: No-U-Turn sampler

In kasparmartens/NUTS: No-U-Turn sampler for HMC

Description

No-U-Turn sampler

Usage

NUTS(theta, f, grad_f, n_iter, M_diag = NULL, M_adapt = 50, delta = 0.5,
  max_treedepth = 10, eps = 1, verbose = TRUE)

Arguments

theta	Initial value for the parameters
f	log-likelihood function (up to a constant)
grad_f	the gradient of the log-likelihood function
n_iter	Number of MCMC iterations
M_diag	Diagonal elements of the mass matrix in HMC. Defaults to ones.
M_adapt	Parameter M_adapt in algorithm 6 in the NUTS paper
delta	Target acceptance ratio, defaults to 0.5
max_treedepth	Maximum depth of the binary trees constructed by NUTS
eps	Starting guess for epsilon

Value

Matrix with the trace of sampled parameters. Each mcmc iteration in rows and parameters in columns.

kasparmartens/NUTS documentation built on May 7, 2019, 11:14 a.m.