1: \begin{abstract}

2: Dimensionality reduction and matrix factorization techniques are important and useful machine-learning techniques in many fields.  

3: \textit{Nonnegative matrix factorization} (NMF) is particularly useful for spectral analysis and image processing in astronomy.

4: I present the vectorized update rules and an independent proof of their convergence for NMF  with heteroscedastic measurements and missing data.

5: I release a Python implementation of the rules and use an optical spectroscopic dataset of extragalactic sources as an example for demonstration.

6: 

7: %to demonstrate that the algorithm and the code can deal with nonuniform uncertainties and missing data in an efficient and robust way.

8: %some of my research notes on NMF, including an alternative proof of the vectorized updated rules  

9: %for NMF with heteroscedastic measurements and missing data. 

10: %along which I release a Python implementation of the updated rules

11: 

12: %This document presents some of my research notes on NMF

13: %I present simple vectorized update rules for \textit{nonnegative matrix factorization} (NMF) with heteroscedastic measurements and missing data,

14: %which can be easily implemented in a modern vector language. 

15: %The sequential multiplicative update rules do not only allow us to construct a set of 

16: %nonnegative basis components given a dataset, but also to decompose a dataset with ease if a basis set is known.

17: \end{abstract}

18: