Privacy Amplification by Structured Subsampling for Deep Differentially Private Time Series Forecasting
Published in International Conference on Machine Learning (ICML), 2025
Structured subsampling for DP time-series; tight event- and user-level guarantees; empirical validation.