A Private and Computationally-Efficient Estimator for Unbounded Gaussians.
Gautam KamathArgyris MouzakisVikrant SinghalThomas SteinkeJonathan R. UllmanPublished in: COLT (2022)
Keyphrases
- computationally efficient
- estimation algorithm
- least squares
- linear combination
- maximum likelihood
- privacy preserving
- error estimation
- gaussian mixture model
- maximum a posteriori
- computational complexity
- conditional expectation
- motion estimator
- data sets
- private data
- public private
- information systems
- gaussian kernels
- variance estimator
- mixture of gaussians
- importance sampling
- private information
- confidence intervals
- estimation error
- mixture model
- model selection
- optical flow
- computer vision