Fit netprioR model

Description

Infer parameters and hidden data using the EM algorithm of netprioR

Usage

1
2
3
4
learn(Yobs, X, G, l, u, a = 0.1, b = 0.1, sigma2 = 1, tau2 = 10,
  eps = 1e-11, max.iter = 500, thresh = 0.001, use.cg = TRUE,
  thresh.cg = 1e-05, nrestarts = 5, max.cores = detectCores(),
  verbose = FALSE)

Arguments

Yobs

Observed labels (NA, if not observed)

X

Phenotypes

G

Graph Laplacians

l

Indices of labelled instances

u

Indices of unlabelled instances

a

Shape parameter of Gamma prior for W

b

Scale parameter of Gamma prior for W

sigma2

Cariance for Gaussian labels

tau2

Variance for Gaussian prior for beta

eps

Small value added to diagonal of Q in order to make it non-singular

max.iter

Maximum number of iterations for EM

thresh

Threshold for termination of EM with respect to change in parameters

use.cg

Flag whether to use conjugate gradient instead of exact computation of expectations

thresh.cg

Threshold for the termination of the conjugate gradient solver

nrestarts

Number of restarts for EM

max.cores

Maximum number of cores to use for parallel computation

verbose

Print verbose output

Value

List containing: Predicted labels Yhat and inferred parameters W and beta

Author(s)

Fabian Schmich