NMFPois: Non-negative matrix factorization algorithm for Poisson data...

In ragnhildlaursen/SFS: Sampling the set of feasible solutions from a non-negative matrix factorization

Description

Non-negative matrix factorization algorithm for Poisson data

Factorizing M into two matrices P and E of dimension ncol(M) x N and N x nrow(M) with the acceleration of SQUAREM. The objective function is the generalized Kullback-Leibler divergence(GKLD).

Usage

NMFPois(M, N, seed = sample(1:100, 3), arrange = TRUE, tol = 1e-05)

Arguments

M	Non-negative data matrix of size
N	Small dimension of the two new matrices
seed	Vector of random seeds to initialize the matrices
arrange	Arranging columns in P and rows in E after largest row sums of E
tol	Maximum change of P and E when stopping

Value

A list of the matrices derived by the factorization and the corresponding generalized Kullback-Leibler

• P - Non-negative matrix of dimension ncol(V) x K, with columns summing to one

• E - Non-negative matrix of dimension K x nrow(V), where rows sum to one

• gkl - Smallest Value of the Generalized Kullback-Leibler

ragnhildlaursen/SFS documentation built on Nov. 15, 2021, 8:28 p.m.