fit_buds: Fit Bayesian Unidimensional Scaling model to

In nlhuong/buds: Bayesian Unidimensional Scaling (BUDS)

Description

Fitting BUDS model to recover latent coordinates of data points from observed/precomputed pairwise dissimilarities

Usage

fit_buds(D, K = NULL, method = c("vb", "mcmc"),
  hyperparams = list(gamma_tau = 2.5, gamma_epsilon = 2.5, gamma_bias =
  2.5, gamma_rho = 2.5, min_sigma = 0.03), init_from = c("random",
  "principal_curve"), seed = 1234, max_trials = 20, ...)

Arguments

D	A square matrix of pairwise dissimilarities
K	An integer indicating k-nearest neighbours to compute average distances to kNN then used as weights in the power likelihood.
method	Stan mode to use either "mcmc" or "vb"
hyperparams	A list of BUDS hiperparameters with named components set by default as follows "gamma_tau"= 2.5, "gamma_epsilon" = 2.5, "gamma_bias" = 2.5, "gamma_rho_sq" = 2.5, "min_sigma" = 0.01
init_from	Initialization option either "random" or from a location along "principal_curve" fitted to 2D representation using PCoA on D, or named list of initialized values for tau (bias, rho, taushape1, taushape2, meansd).
seed	A integer representing a random seed
max_trials	A integer for maximum number of trials to fit data
...	Other parameters for vb and stan

Value

The fitted 'stanfit' model object

nlhuong/buds documentation built on May 17, 2019, 3:13 a.m.