The paper shows that if a function class is rho-separated, a Gaussian-perturbed follow-the-leader algorithm achieves small-loss regret and differentially private learning rates using an ERM oracle.
Learning in non-convex games with an optimization oracle
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Small Loss Bounds for Online Learning Separated Function Classes: A Gaussian Process Perspective
The paper shows that if a function class is rho-separated, a Gaussian-perturbed follow-the-leader algorithm achieves small-loss regret and differentially private learning rates using an ERM oracle.