Kernel learning at the first level of inference
- Submitting institution
-
The University of East Anglia
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 182619719
- Type
- D - Journal article
- DOI
-
10.1016/j.neunet.2014.01.011
- Title of journal
- Neural Networks
- Article number
- -
- First page
- 69
- Volume
- 53
- Issue
- -
- ISSN
- 0893-6080
- Open access status
- Out of scope for open access requirements
- Month of publication
- May
- Year of publication
- 2014
- URL
-
http://www.scopus.com/inward/record.url?scp=84894104405&partnerID=8YFLogxK
- 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
- 2
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- This paper demonstrates that while kernel learning methods provide a reliable means of avoiding over-fitting in the estimation of the coefficients of the kernel expansion, the problem of over-fitting has largely been merely shifted to model selection, where the parameters of the kernel function are tuned, rather than actually solved. To more fully solve the problem of over-fitting, we therefore need to extend the theory and algorithms to tuning the kernel. This paper gives a simple approach, that is significantly better than traditional approaches and competitive with the state-of-the-art, for rich kernels with many degrees of freedom.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -