Differentially Private Regression with Gaussian Processes
- Submitting institution
-
The University of Sheffield
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 5650
- Type
- E - Conference contribution
- DOI
-
-
- Title of conference / published proceedings
- Proceedings of the 21st International Conference on Artificial Intelligence and Statistics
- First page
- 1195
- Volume
- 84
- Issue
- -
- ISSN
- 2640-3498
- Open access status
- Access exception
- Month of publication
- March
- Year of publication
- 2018
- 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)
-
C - Machine Learning
- Citation count
- 1
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- Applying differential privacy (DP) often leads to excessive added noise. The first paper to combine DP with Gaussian processes to address this issue. Our technique allows predictions to remain useful even from only a few data. Though recently published, it has already been extended to provide DP GP classification and made even more robust with a sparse approximation by author and collaborators (https://arxiv.org/abs/1909.09147). Used in several papers in the field (http://proceedings.mlr.press/v97/mirshani19a.html, doi.org/10.1007%2F978-3-030-27615-7_14, http://proceedings.mlr.press/v97/awan19a.html) and cited in wider literature, e.g. anomaly detection (http://proceedings.mlr.press/v71/adelsberg18a.html). Supported by the EPSRC Research Project EP/N014162/1.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -