volnmf_simplex_col: Update of a matrix in NMF with equality contstraints on...

In vrnmf: Volume-Regularized Structured Matrix Factorization

Update of a matrix in NMF with equality contstraints on columns.

Description

volnmf_simplex_col finds non-negative matrix C that minimizes the objective ||X-C*R||^2 under constraints that columns of C equal to 1 using local approximation with extrapolation.

Usage

volnmf_simplex_col(
  X,
  R,
  C.prev = NULL,
  bound = 1,
  extrapolate = TRUE,
  err.cut = 1e-10,
  n.iter = 10000,
  qmax = 100
)

Arguments

X	Numeric Matrices. Matrices involved in the objective function.
R	Numeric Matrices. Matrices involved in the objective function.
C.prev	Numeric Matrices. Matrices involved in the objective function. Matrix C.prev serves as initialization. (default=NULL)
bound	A numeric. Equality constraint on columns of matrix C. (default=1)
extrapolate	A boolean. Use extrapolation after local approximation. (default=TRUE)
err.cut	A numeric. Stop iterations if relative error between iterations is less than err.cut (parameter is not active now). (default=1e-10)
n.iter	An integer. Number of iterations. (default=1000)
qmax	A numeric. Maximum asymptotic (1 - 1/qmax) of extrapolation step.

Value

An updated matrix C.

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