The aim of this post is to highlight the utility of non-negative factorization (NMF) in data analysis through examples. In a different post, we’ll talk theory and implementation by thinking of NMF as constrained optimization.

**Learning parts of a whole**

When Lee and Seung [Lee99] re-introduced non-negative matrix factorization (NMF), they emphasized on the algorithm’s ability to learn parts of a whole. Continue reading