> ML_LITERATURE // ABADI-2016-DEEP-LEARNING-WITH-DIFFERENTIAL-PRIVACY-DPSGD_v1.0
Deep Learning with Differential Privacy (DP-SGD)
Martin Abadi, Andy Chu, Ian Goodfellow, H. Brendan McMahan, Ilya Mironov, Kunal Talwar, Li Zhang · ACM SIGSAC Conference on Computer and Communications Security (CCS) (2016)
safety-fairness2016industry-standardartifactsAvailable
Principal Contribution
Invented Differentially Private Stochastic Gradient Descent (DP-SGD), clipping per-sample gradients and injecting calibrated Gaussian noise tracked via the Moments Accountant.
Operational Relevance
Serves as qualified reference for implementing task-privacy-preserving-ml in production systems.
Assumptions
- Underlying computational topology and mathematical bounds adhere to established convexity/smoothness guarantees
Limitations
- Hardware runtime speedups, privacy budgets, and convergence depend on hyperparameters and network communication limits
Connected Algorithms, Architectures & Tools
Related Algorithms:
Related Architectures:
Implementing Libraries: