predict.gmf: Predict generalized PCA scores or reconstruction on new data

In andland/generalizedPCA: Generalized PCA

Description

Predict generalized PCA scores or reconstruction on new data

Usage

## S3 method for class 'gmf'
predict(object, newdata, type = c("PCs", "link", "response"),
  quiet = TRUE, max_iters = 1000, conv_criteria = 1e-05, start_A, ...)

Arguments

object	generalized MF object
newdata	matrix of the same exponential family as covariates in object. If missing, will use the data that object was fit on
type	the type of fitting required. type = "PCs" gives matrix of principal components of x, type = "link" gives a matrix on the natural parameter scale, and type = "response" gives a matrix on the response scale
quiet	logical; whether the calculation should show progress
max_iters	maximum number of iterations
conv_criteria	convergence criteria
start_A	initial value for A
...	Additional arguments

Examples

# construct a low rank matrices in the natural parameter space
rows = 100
cols = 10
set.seed(1)
loadings = rnorm(cols)
mat_np = outer(rnorm(rows), rnorm(cols))
mat_np_new = outer(rnorm(rows), loadings)

# generate a count matrices
mat = matrix(rpois(rows * cols, c(exp(mat_np))), rows, cols)
mat_new = matrix(rpois(rows * cols, c(exp(mat_np_new))), rows, cols)

# run Poisson PCA on it
gmf = generalizedMF(mat, k = 1, family = "poisson")

A = predict(gmf, mat_new)

andland/generalizedPCA documentation built on May 12, 2019, 2:42 a.m.