Shuyan Zou

SlaClip: Gradient Norm Slacks can be Indicator for Adaptive Clipping in DP-SGD

Differentially private stochastic gradient descent (DP-SGD) achieves privacy by clipping per-sample gradients and injecting Gaussian noise, but its utility is highly sensitive to the choice of the clipping threshold $C$. A fixed $C$ often degrades performance and necessitates repeated empirical cali…