Private Realizable-to-Agnostic Transformation with Near-Optimal Sample Complexity
This work improves sample efficiency for private agnostic learning, benefiting privacy-preserving machine learning by addressing a bottleneck in existing transformations.
The paper tackles the problem of converting private learners from realizable to agnostic settings with suboptimal sample complexity due to privacy parameters, achieving a near-optimal extra sample complexity of O~(VC(C)/α²) for any ε ≤ 1, and applies this to resolve an open question in private prediction.
The realizable-to-agnostic transformation (Beimel et al., 2015; Alon et al., 2020) provides a general mechanism to convert a private learner in the realizable setting (where the examples are labeled by some function in the concept class) to a private learner in the agnostic setting (where no assumptions are imposed on the data). Specifically, for any concept class $\mathcal{C}$ and error parameter $α$, a private realizable learner for $\mathcal{C}$ can be transformed into a private agnostic learner while only increasing the sample complexity by $\widetilde{O}(\mathrm{VC}(\mathcal{C})/α^2)$, which is essentially tight assuming a constant privacy parameter $\varepsilon = Θ(1)$. However, when $\varepsilon$ can be arbitrary, one has to apply the standard privacy-amplification-by-subsampling technique (Kasiviswanathan et al., 2011), resulting in a suboptimal extra sample complexity of $\widetilde{O}(\mathrm{VC}(\mathcal{C})/α^2\varepsilon)$ that involves a $1/\varepsilon$ factor. In this work, we give an improved construction that eliminates the dependence on $\varepsilon$, thereby achieving a near-optimal extra sample complexity of $\widetilde{O}(\mathrm{VC}(\mathcal{C})/α^2)$ for any $\varepsilon\le 1$. Moreover, our result reveals that in private agnostic learning, the privacy cost is only significant for the realizable part. We also leverage our technique to obtain a nearly tight sample complexity bound for the private prediction problem, resolving an open question posed by Dwork and Feldman (2018) and Dagan and Feldman (2020).