Security, privacy, and fairness · MLD-0337
DP-SGD
SGD with per-example or per-unit gradient clipping and calibrated noise, tracked with privacy accounting.
Original reference page · detailed expansion pending. Original definitions, examples, notes and citations are preserved below. New examples are illustrative, not benchmark results.
Original example
Train with a formal privacy guarantee under assumptions.
Original caveat
Ordinary gradient noise or clipping alone is not sufficient.
Context and aliases
Level: Intermediate. Stages: Data · Training · Evaluation · Deployment · Governance.
Tags: ML · DL · SLM · LLM · Security · Privacy · Safety.
Aliases: Differentially private stochastic gradient descent.
Related concepts
Concepts linking here
Original references
- The Algorithmic Foundations of Differential Privacy ↗ · Reference link
- Google Machine Learning Glossary ↗ · Consulted
- OWASP Top 10 for LLM Applications ↗ · Reference link
- Original reference URL ↗