Dictionary Learning with BLOTLESS Update
- Submitting institution
-
King's College London
- Unit of assessment
- 12 - Engineering
- Output identifier
- 123658941
- Type
- D - Journal article
- DOI
-
10.1109/TSP.2020.2971948
- Title of journal
- IEEE Transactions on Signal Processing
- Article number
- 8985423
- First page
- 1635
- Volume
- 68
- Issue
- -
- ISSN
- 1053-587X
- Open access status
- Compliant
- Month of publication
- February
- Year of publication
- 2020
- 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
-
3
- Research group(s)
-
-
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- Learning a dictionary under which given data can be represented sparsely has been one of central problems in signal processing over the past two decades, as it is a key ingredient of compressed sensing, enabling a broad range of associated applications (e.g. image and video processing, denoising, physiological signal analysis). Whilst scores of heuristic dictionary learning algorithms have been proposed, guaranteed performance and fundamental theoretical underpinnings have been lacking. This paper establishes three strong theoretical results on performance guarantees and proposes an algorithm which outperforms state of the art by a large margin especially in scenarios with scarce training data.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -