Nonnegative Matrix Factorization

NMF is a python program that applies a choice of nonnegative matrix factorization (NMF) algorithms to a dataset for clustering.
Currently, this program supports

• Multiplicative Updates (MU)1
• Alternating Least Squares (ALS)2
• Alternating Nonnegative Least Squares with Active Set (ANLS - AS)3

Results

Experimental results with abcnews-date-test.csv's headline_text
Multiplicative Updates (MU):

Alternating Least Squares (ALS):

How to Run

• Usage: main.py [-h] -f FILENAME -c COL_NAME -m {sklearn,all,als,anls,mu} [-d DATA_FRAC] [-r RANDOM_SAMPLE] [-n NUM_MAX_FEATURE] [-s CLUSTER_SIZE] [-k NUM_CLUSTERS] [-i NUM_ITERS] [-p PRINT_ENABLED]

• Required arguments:

  • -f FILENAME, --filename FILENAME
    the input file name
  • -c COL_NAME, --col_name COL_NAME
    the column of the input csv file for nonnegative matrix factorization.
  • -m {sklearn,all,als,anls_as,mu}, --method {sklearn,all,als,anls_as,mu}
    the NMF method to apply

• Optional arguments:

  • -h, --help
    show this help message and exit
  • -d DATA_FRAC, --data_frac DATA_FRAC
    the amount of the data to be used
  • -r RANDOM_SAMPLE, --random_sample RANDOM_SAMPLE
    if set False, disables random sampling of the data
  • -n NUM_MAX_FEATURE, --num_max_feature NUM_MAX_FEATURE
    the maximum number of features to be discovered in the dataset
  • -s CLUSTER_SIZE, --cluster_size CLUSTER_SIZE
    the number of features in each cluster
  • -k NUM_CLUSTERS, --num_clusters NUM_CLUSTERS
    the number of clusters to be discovered
  • -i NUM_ITERS, --num_iters NUM_ITERS
    the number of iterations to run a NMF algorithm
  • -p PRINT_ENABLED, --print_enabled PRINT_ENABLED
    if ture, output print statements

Citation

Algorithms for Non-negative Matrix Factorizations by D. Lee and H. Seung,
https://papers.nips.cc/paper/1861-algorithms-for-non-negative-matrix-factorization.pdf
Algorithms and applications for approximate nonnegative matrix factorization by M. Berry,
https://www.sciencedirect.com/science/article/pii/S0167947306004191
Non-negative Matrix Factorization Based on Alternating Non-negativity Constrained Least Squares and ActiveSet Method by H. Kim and H. Park,
https://www.cc.gatech.edu/~hpark/papers/simax-nmf.pdf
Fast Nonnegative Matrix Factorization: An Active-Set-Like Method and Comparisons,
https://www.cc.gatech.edu/~hpark/papers/SISC_082117RR_Kim_Park.pdf