rsim_data: Simulate a data matrix of a given dimension and rank for PCP...

In Columbia-PRIME/PCPhelpers: Provides a bunch of functions to help with PCP experimentation.

Simulate a data matrix of a given dimension and rank for PCP experiments.

Description

rsim_data simulates a data matrix, M, of a given dimension and rank by simulating ground truth low rank (L), sparse (S) and noise (Z) components. Meaning M = L + S + Z. The M, L, S and Z matrices are all accessible from this function, allowing for easy evaluation of PCP's performance.

Usage

rsim_data(
  sim_seed,
  nrow,
  ncol,
  rank,
  sigma,
  add_sparse = FALSE,
  nonneg = TRUE,
  ldist = runif,
  sdist = runif,
  zdist = rnorm,
  ...
)

Arguments

sim_seed	Required seed to allow for reproducible results & distinct simulated matrices.
nrow	Sets number of rows in the simulated data matrix.
ncol	Sets the number of columns in the simulated data matrix.
rank	Sets the rank of the simulated data matrix.
sigma	Sets the standard deviation of the noise matrix, Z.
add_sparse	Logical that sets whether sparse noise, S, is added or not. Default=FALSE.
nonneg	Logical that sets whether the final M matrix should be non-negative. Default=TRUE.
ldist	Function that defines a distribution to use when simulating the L matrix. Default=runif.
sdist	Function that defines a distribution to use when simulating the S matrix. Default=runif.
zdist	Function that defines a distribution to use when simulating the Z matrix. Default=rnorm.

Value

list containing the simulated M, L, S and Z matrices.

Examples

data <- rsim_data(sim_seed = 1, nrow = 10, ncol = 10, rank = 3, sigma = .1, add_sparse = TRUE)

Columbia-PRIME/PCPhelpers documentation built on April 24, 2022, 7:57 p.m.