Auto DP-SGD: Dual Improvements of Privacy and Accuracy via Automatic Clipping Threshold and Noise Multiplier Estimation.
Sai Venkatesh ChilukotiMd. Imran HossenLiqun ShanVijay Srinivas TidaXiai HeiPublished in: CoRR (2023)
Keyphrases
- estimation error
- estimation accuracy
- high accuracy
- dynamic programming
- prediction accuracy
- privacy preserving
- error analysis
- window size
- noise reduction
- missing data
- estimated parameters
- signal to noise ratio
- measurement error
- noise level
- data mining
- stochastic gradient descent
- noise immunity
- private information
- personal information
- third party
- error rate
- least squares
- computational cost
- computational complexity