Condition estimation for regression and feature selection
- Submitting institution
-
The University of Sheffield
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 2515
- Type
- D - Journal article
- DOI
-
10.1016/j.cam.2019.03.041
- Title of journal
- Journal of Computational and Applied Mathematics
- Article number
- 112212
- First page
- -
- Volume
- 373
- Issue
- -
- ISSN
- 0377-0427
- Open access status
- Exception within 3 months of publication
- Month of publication
- April
- Year of publication
- 2019
- 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
-
1
- Research group(s)
-
A - Algorithms
- Citation count
- 0
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- Many problems in machine learning (ML) yield an unstable linear equation, which is stabilised by regularisation. Regularisation is well-understood mathematically but is often used incorrectly in ML because the criterion for its correct use is not established. This criterion leads to a distinction between the stability of forward and inverse problems, which is important in ML. Led to preliminary discussions with Hansen (Mathematics, Cambridge) exploring key mathematical problems in deep learning (LMS-Bath Symposium 2020, Hansen). The result also has applications to compressed sensing, which allows sampling below the Nyquist rate, making it attractive in medical imaging.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -