NNMF_from_file: Nonnegative matrix factorization

In KSheu/ksheu.library1: KSheu Library1

Nonnegative matrix factorization

Description

Input: matrix A (ex. rows as genes, cols as samples) Output: writes to file decomposed matrices W (loadings) and H (scores) that minimize the loss between the original and decomposed matrices

Usage

NNMF_from_file(file, rank = 5, method = "scd", loss = "mkl",
  impute = F, verbose = 2)

Arguments

file	file for A matrix (.txt or .RDS file)
rank	rank for the decompostiton
method	'scd' (seq. coord-wise descent) or 'lee' (Lee's algo)
loss	'mse' (mean squared error) or 'mkl' (mean KL divergence) for sc data probably better to use KL divergence? (Frobenius norm is standard)
impute	default=F Simple NNMF or impute 0 values by first setting them all to NA
verbose	0-2

Details

Returns nnmf object

KSheu/ksheu.library1 documentation built on Jan. 28, 2025, 3:26 p.m.