Per-Instance Privacy Accounting for Differentially Private Stochastic Gradient Descent.
Da YuGautam KamathJanardhan KulkarniTie-Yan LiuJian YinHuishuai ZhangPublished in: CoRR (2022)
Keyphrases
- differentially private
- stochastic gradient descent
- differential privacy
- least squares
- matrix factorization
- loss function
- step size
- privacy preserving
- privacy guarantees
- support vector machine
- private data
- random forests
- regularization parameter
- multiple kernel learning
- importance sampling
- weight vector
- support vector
- collaborative filtering
- cost function
- data sets
- recommender systems
- online algorithms
- lower bound