Variational Sparse Coding
- Submitting institution
-
University of Glasgow
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 11-04711
- Type
- E - Conference contribution
- DOI
-
-
- Title of conference / published proceedings
- Conference on Uncertainty in Artificial Intelligence (UAI 2019)
- First page
- 1
- Volume
- -
- Issue
- -
- ISSN
- -
- Open access status
- -
- Month of publication
- July
- Year of publication
- 2019
- URL
-
http://eprints.gla.ac.uk/191553/
- 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
-
2
- Research group(s)
-
-
- Citation count
- -
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- ORIGINALITY: presents the first algorithm which can robustly estimate sparse and interpretable latent factors in non-linear and high-dimensional datasets without supervision. Interpretable models are a prerequisite for the success of safe AI, and the developed algorithm provides a general approach for learning interpretable and generative representations in high-dimensional data with application in for example, exploratory medical image analysis. SIGNIFICANCE: published at a top machine learning conference. RIGOUR: algorithm is derived mathematically. The method is evaluated on two defacto benchmarks and a custom problem demonstrating its superiority over alternative models in finding sparse and easily interpretable representations.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -