Compressive learning with privacy guarantees. (15th May 2021)
- Record Type:
- Journal Article
- Title:
- Compressive learning with privacy guarantees. (15th May 2021)
- Main Title:
- Compressive learning with privacy guarantees
- Authors:
- Chatalic, A
Schellekens, V
Houssiau, F
de Montjoye, Y A
Jacques, L
Gribonval, R - Abstract:
- Abstract: This work addresses the problem of learning from large collections of data with privacy guarantees. The compressive learning framework proposes to deal with the large scale of datasets by compressing them into a single vector of generalized random moments, called a sketch vector, from which the learning task is then performed. We provide sharp bounds on the so-called sensitivity of this sketching mechanism. This allows us to leverage standard techniques to ensure differential privacy—a well-established formalism for defining and quantifying the privacy of a random mechanism—by adding Laplace of Gaussian noise to the sketch. We combine these standard mechanisms with a new feature subsampling mechanism, which reduces the computational cost without damaging privacy. The overall framework is applied to the tasks of Gaussian modeling, k-means clustering and principal component analysis, for which sharp privacy bounds are derived. Empirically, the quality (for subsequent learning) of the compressed representation produced by our mechanism is strongly related with the induced noise level, for which we give analytical expressions.
- Is Part Of:
- Information and inference. Volume 11:Number 1(2022)
- Journal:
- Information and inference
- Issue:
- Volume 11:Number 1(2022)
- Issue Display:
- Volume 11, Issue 1 (2022)
- Year:
- 2022
- Volume:
- 11
- Issue:
- 1
- Issue Sort Value:
- 2022-0011-0001-0000
- Page Start:
- 251
- Page End:
- 305
- Publication Date:
- 2021-05-15
- Subjects:
- compressive learning -- privacy-aware learning -- differential privacy -- sketching
Mathematical models -- Periodicals
519.605 - Journal URLs:
- http://imaiai.oxfordjournals.org/ ↗
http://ukcatalogue.oup.com/ ↗ - DOI:
- 10.1093/imaiai/iaab005 ↗
- Languages:
- English
- ISSNs:
- 2049-8764
- Deposit Type:
- Legaldeposit
- View Content:
- Available online (eLD content is only available in our Reading Rooms) ↗
- Physical Locations:
- British Library DSC - BLDSS-3PM
British Library HMNTS - ELD Digital store - Ingest File:
- 21338.xml