Learning kernel logistic regression in the presence of class label noise
- Submitting institution
-
The University of Birmingham
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 42748341
- Type
- D - Journal article
- DOI
-
10.1016/j.patcog.2014.05.007
- Title of journal
- Pattern Recognition
- Article number
- -
- First page
- 3641
- Volume
- 47
- Issue
- 11
- ISSN
- 0031-3203
- Open access status
- Out of scope for open access requirements
- Month of publication
- May
- Year of publication
- 2014
- 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)
-
-
- Citation count
- 28
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- A principled probabilistic modelling approach to account for labelling
errors in nonlinear classification, including a method to set model
complexity parameters without availability of a trusted validation
set. A thorough empirical testing and two real-world applications
demonstrate the success of the approach. This work led to guest
editing a special issue of Neurocomputing dedicated to the topic
(Frenay & Kaban, 2015), speaking invitations at LABELNOISE'17
(https://labelnoise2017.loria.fr/people/) and at special sessions of
EcoStat'18 and CMStat'18. It motivated in-depth theoretical research
(Reeve & Kaban, COLT 2019), (Reeve & Kaban, ICML 2019) that led to a
Statistics lectureship at Bristol University for Reeve.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -