Verified75% confidenceFactExact time
DP-SGD is a mechanism for achieving differential privacy in stochastic gradient descent, implemented by clipping per-sample gradients and then adding Gaussian noise.
3
Sources
75%
Confidence
Long-term
Relevance
10/13/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedStochastic gradient descent (SGD) uses only one sample at a time, which is fast but produces noisy update directions.77% similarUnverified训练时应用差分隐私噪声（DP-SGD）可有效降低记忆化程度，但会带来个性化效果的损耗76% similarUnverifiedExtending the AMP framework from full-batch gradient descent to stochastic gradient descent (SGD) requires handling the additional sampling randomness and data reuse bias.75% similarUnverifiedStochastic gradient descent (SGD) is a cornerstone of machine learning optimization, and rigorous analysis of its convergence rate usually relies on cumbersome probabilistic derivations such as martingale theory or concentration inequalities.75% similarUnverifiedFine-tuning via one-pass SGD also achieves the same optimal rate.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1020473API
curl https://kongchang.com/api/v1/knowledge/claims/1020473MCP
get_claim(id=1020473)