Rank selection in non-negative matrix factorization using minimum description Length
- Submitting institution
-
University of Southampton
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 49846429
- Type
- D - Journal article
- DOI
-
10.1162/NECO_a_00980
- Title of journal
- Neural Computation
- Article number
- -
- First page
- 2164
- Volume
- 29
- Issue
- 8
- ISSN
- 1530-888X
- Open access status
- Compliant
- Month of publication
- July
- Year of publication
- 2017
- URL
-
-
- Supplementary information
-
-
- Request cross-referral to
- -
- Output has been delayed by COVID-19
- No
- COVID-19 affected output statement
- -
- Forensic science
- No
- Criminology
- No
- Interdisciplinary
- No
- Number of additional authors
-
2
- Research group(s)
-
-
- Citation count
- 10
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- This paper is representative of a body of work published in 5 papers (https://tinyurl.com/wuhxbpr) on applying minimum description length techniques to non-negative matrix factorisation. This includes an application of the work described in this paper to finance. The paper provides a rational approach to selecting the rank and other hyperparameters using the data itself. This provides an alternative to the cumbersome and often unpractical process of cross-validation, which, as shown in the paper, actually rarely works.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -