Symmetric matrix perturbation for differentially-private principal component analysis.
Hafiz ImtiazAnand D. SarwatePublished in: ICASSP (2016)
Keyphrases
- differentially private
- symmetric matrix
- covariance matrix
- principal component analysis
- differential privacy
- privacy guarantees
- eigenvalue decomposition
- principal components
- independent component analysis
- semidefinite programming
- low dimensional
- dimensionality reduction
- linear discriminant analysis
- feature extraction
- dimension reduction
- sample size
- face recognition
- semidefinite
- transformation matrix
- canonical form
- privacy preserving
- probabilistic model
- random projections
- feature space
- data sets
- information retrieval