sim_factors: Simulate matrices to explores 'vrnmf'

In vrnmf: Volume-Regularized Structured Matrix Factorization

Simulate matrices to explores vrnmf

Description

sim_factors simulates non-negative factorization matrices C and D under a variaty of conditions to explore factorization X = C*D + noise.

Usage

sim_factors(
  m,
  n,
  r,
  simplex = "col",
  distr = "unif",
  frac.zeros = 0.4,
  condition = FALSE,
  noise = 0
)

Arguments

m	Integers. Size of matrices. Matrix C has a size of m*r and matrix D has a size of r*n.
n	Integers. Size of matrices. Matrix C has a size of m*r and matrix D has a size of r*n.
r	Integers. Size of matrices. Matrix C has a size of m*r and matrix D has a size of r*n.
simplex	A character. Either columns ("col") or rows ("row") of matrix C are projected onto unit simplex. (default="col")
distr	A character. Distribution to simulate matrix entries: "unif" for uniform and "exp" for exponential distributions. (default="unif")
frac.zeros	A numeric. Fraction of zeros in matrix C. It promotes sufficient scattering of matrix column/row vectors. (default=0.4)
condition	A boolean. Generate more well-conditioned matrix R. (default=FALSE)
noise	A numeric. Standard deviation of gaussian noise to add. (default=0e-4)

Value

List of simulated matrices:

X.noise, X - noisy and original matrix X to decompose.

C, D - factorization matrices.

vrnmf documentation built on March 18, 2022, 6:11 p.m.