Oracle Efficient Private Non-Convex Optimization

TL;DR

This paper introduces two new differentially private optimization algorithms that work without assuming convexity, applicable to both discrete and continuous domains, and demonstrates their effectiveness in non-convex learning tasks.

Contribution

It presents two oracle-efficient private optimization algorithms that extend objective perturbation to non-convex settings over discrete and continuous domains.

Findings

Algorithms work without convexity assumptions.

Empirical results show improved performance in non-convex learning.

Applicable to learning linear classifiers with non-convex loss functions.

Abstract

One of the most effective algorithms for differentially private learning and optimization is objective perturbation. This technique augments a given optimization problem (e.g. deriving from an ERM problem) with a random linear term, and then exactly solves it. However, to date, analyses of this approach crucially rely on the convexity and smoothness of the objective function, limiting its generality. We give two algorithms that extend this approach substantially. The first algorithm requires nothing except boundedness of the loss function, and operates over a discrete domain. Its privacy and accuracy guarantees hold even without assuming convexity. This gives an oracle-efficient optimization algorithm over arbitrary discrete domains that is comparable in its generality to the exponential mechanism. The second algorithm operates over a continuous domain and requires only that the loss function be bounded and Lipschitz in its continuous parameter. Its privacy analysis does not require convexity. Its accuracy analysis does require convexity, but does not require second order conditions like smoothness. Even without convexity, this algorithm can be generically used as an oracle-efficient optimization algorithm, with accuracy evaluated empirically. We complement our theoretical results with an empirical evaluation of the non-convex case, in which we use an integer program solver as our optimization oracle. We find that for the problem of learning linear classifiers, directly optimizing for 0/1 loss using our approach can out-perform the more standard approach of privately optimizing a convex-surrogate loss function on the Adult dataset.