A random-projection differentially private kernel ERM method attains minimax-optimal excess risk bounds for squared and Lipschitz-smooth convex losses under local strong convexity, plus the first dimension-free bounds for objective-perturbation private linear ERM.
Rademacher and gaussian complexities: Risk bounds and structural results
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
verdicts
UNVERDICTED 2representative citing papers
Generalization and learnability bounds for hypothesis classes under Dobrushin's condition on weakly dependent data, with degradation by only constant or log factors relative to i.i.d. settings.
citing papers explorer
-
Optimal differentially private kernel learning with random projection
A random-projection differentially private kernel ERM method attains minimax-optimal excess risk bounds for squared and Lipschitz-smooth convex losses under local strong convexity, plus the first dimension-free bounds for objective-perturbation private linear ERM.
-
Learning from weakly dependent data under Dobrushin's condition
Generalization and learnability bounds for hypothesis classes under Dobrushin's condition on weakly dependent data, with degradation by only constant or log factors relative to i.i.d. settings.