Infinite feature selection: a graph-based feature filtering approach
- Submitting institution
-
University of Glasgow
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 11-12024
- Type
- D - Journal article
- DOI
-
10.1109/TPAMI.2020.3002843
- Title of journal
- IEEE Transactions on Pattern Analysis and Machine Intelligence
- Article number
- -
- First page
- 1
- Volume
- 0
- Issue
- -
- ISSN
- 0162-8828
- Open access status
- Compliant
- Month of publication
- June
- Year of publication
- 2020
- URL
-
http://eprints.gla.ac.uk/218830/
- 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
-
4
- Research group(s)
-
-
- Citation count
- -
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- ORIGINALITY: New approach that performs feature selection for machine learning by identifying a subpath, possibly of infinite length, in a graph. RIGOUR: Algorithm is systematically compared with 18 state-of-the-art feature selection methodologies, over 11 of the most important publicly available benchmarks to measure performance. The algorithm performs significantly better than the other 18 methodologies over most benchmarks (and never less than the others). SIGNIFICANCE: Published in top IEEE journal in pattern analysis and machine learning. Builds upon earlier work cited 130+ times since 2017 and published at IEEE ICCV, the top computer vision conference (https://ieeexplore.ieee.org/document/8237418).
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -
