FindNumFactors: Determines the optimal number of NMF factors to use by...

In yanwu2014/swne: Similarity Weighted Nonnegative Embedding: A method for visualizing high dimensional datasets

Determines the optimal number of NMF factors to use by comparing the reduction in reconstruction error vs the reduction in reconstruction error for a randomized matrix. The optimal number of factors is where the decrease in reconstruction errors intersect

Description

Adapted from Frigyesi et al, 2008. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2623306/

Usage

FindNumFactors(
  A,
  k.range = seq(2, 12, 2),
  n.cores = 1,
  do.plot = T,
  seed = NULL,
  loss = "mse",
  max.iter = 250
)

Arguments

A	Input data matrix
k.range	Range of NMF factors to fit over
n.cores	Number of threads
do.plot	Whether to plot the reconstruction error
seed	Random seed for randomizing matrix
loss	Loss function to use for NMF
max.iter	Maximum iterations for NMF run

Value

Reconstruction error at each number of NMF factors specified in k.range

yanwu2014/swne documentation built on Aug. 5, 2023, 4:42 a.m.