In this paper, we revisit the problem of Differentially Private Stochastic Convex Optimization (DP-SCO) in Euclidean and general `dp spaces. Specifically, we focus on three settings that are still far from well understood: (1) DP-SCO over a constrained and bounded (convex) set in Euclidean space; (2) unconstrained DP-SCO in `dp space; (3) DP-SCO with heavy-tailed data over a constrained and bounded set in `dp space. For problem (1), for both convex and strongly convex loss functions, we propose methods whose outputs could achieve (expected) excess population risks that are only dependent on the Gaussian width of the constraint set, rather than the dimension of the space. Moreover, we also show the bound for strongly convex functions is optimal up to a logarithmic factor. For problems (2) and (3), we propose several novel algorithms and provide the first theoretical results for both cases when 1 < p < 2 and 2 ≤ p ≤ ∞.
|Original language||English (US)|
|Title of host publication||39th Conference on Uncertainty in Artificial Intelligence, UAI 2023|
|Publisher||ML Research Press|
|Number of pages||10|
|State||Published - Mar 31 2023|
Bibliographical noteKAUST Repository Item: Exported on 2023-09-15
Acknowledged KAUST grant number(s): BAS/1/1689-01-01, FCC/1/1976-49-01, URF/1/4663-01-01
Acknowledgements: Di Wang was supported in part by the baseline funding BAS/1/1689-01-01, funding from the CRG grand URF/1/4663-01-01, FCC/1/1976-49-01 from CBRC. He was also supported by the funding of the SDAIA-KAUST Center of Excellence in Data Science and Artificial Intelligence (SDAIA-KAUST AI). Changhong Zhao was supported in part by Hong Kong Research Grants Council through ECS Grant 24210220.