Non-negative Matrix Factorization (NMF) — when to use?
mediumAnswer
- Factor X ≈ WH with W, H ≥ 0.
- Enforces additive parts-based representation.
- Uses: (1) topic modeling on TF-IDF (topics as sparse non-negative bases), (2) audio spectrogram decomposition, (3) image parts (Lee & Seung's faces).
- More interpretable than PCA when non-negativity is natural.
- Fit by multiplicative updates or ADMM.
- Rank must be chosen (CV / stability).
Check yourself — multiple choice
- Random
- X ≈ WH with W,H ≥ 0 → interpretable parts-based decomposition (topics, image parts, audio spectrogram); choose rank via CV
- Same as PCA
- Not real
NMF: non-negative parts-based factorization; topics / audio / images.
#dimensionality-reduction
Practise Unsupervised Learning
214 interview questions in this topic.