Performing non-negative matrix factorization solving argmin_(A >= 0, S >= 0) 1 / 2 ||Y - AS||_2^2 + lambda ||S||_1

nGMCA(originalMatrix: any, rank: Number, options: Object)
Parameters
originalMatrix (any)
rank (Number) The maximum number of linearly independent column/row vectors in the matrix.
options (Object = {}) Options of ngmca factorization method.
Name Description
options.maximumIteration Number (default 500) Maximum number of iterations.
options.maxFBIteration Number (default 80) Maximum number of iterations of the Forward-Backward subroutine.
options.randGenerator Object (default Math.random) Random number generator for the subroutine of initialization.
options.maxInitFBIteration Number (default 50) Maximum number of iterations of the Forward-Backward subroutine at the initialization.
options.toleranceFB Number (default 1e-5) relative difference tolerance for convergence of the Forward-Backward sub-iterations.
options.toleranceFBInit Number (default 0) relative difference tolerance for convergence of the Forward-Backward sub-iterations at the initialization.
options.phaseRatio Number (default 0.8) transition between decreasing thresholding phase and refinement phase in percent of the iterations.
options.tauMAD Number (default 1) constant coefficient for the final threshold computation.
options.useTranspose Boolean (default false) if true the originalMatrix is transposed.