Skip to main content

Table 2 Computational cost of the NMF multiplicative-update algorithm.

From: Discovering gene functional relationships using FAUN (Feature Annotation Using Nonnegative matrix factorization)

NMF Rank k Number of iterations CPU Time (seconds) Millions of ops. per iteration
  50TG BGM NatRev 50TG BGM NatRev 50TG BGM NatRev
10 86 130 92 3.57 14.40 8.20 4 13 14
20 114 133 106 11.55 28.96 15.88 8 26 29
30 162 154 119 28.34 46.57 23.91 13 39 43
40 166 165 147 36.27 70.40 40.59 18 51 57
50 634 171 180 197.96 94.70 61.33 22 64 72
  1. The number of terms (m) for the 50TG, BGM, and NatRev datasets are 8, 750, 12, 590, and 13, 038, respectively. The number of gene documents (n) for the 50TG, BGM, and NatRev datasets are 50, 102, and 110, respectively.