Differentially Private Learning of Structured Discrete Distributions
- Submitting institution
-
University of Edinburgh
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 108309895
- Type
- E - Conference contribution
- DOI
-
-
- Title of conference / published proceedings
- Advances in Neural Information Processing Systems 28
- First page
- 2566
- Volume
- 28
- Issue
- -
- ISSN
- 1049-5258
- Open access status
- Out of scope for open access requirements
- Month of publication
- December
- Year of publication
- 2015
- 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
-
2
- Research group(s)
-
C - Foundations of Computation
- Citation count
- 0
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- The paper introduces an efficient algorithm for differentially private approximations to a large family of distributions. The cost of privacy is only an error sublogarithmic in the domain size. The paper also presents a new heuristic which gives the new algorithm good empirical performance even on input sizes beyond the reach of previous algorithms. The paper appeared in Neurips 2015, which had a 22% acceptance rate.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -