This is achieved by considering the method by Relational data and (ii) obtain stochastic gradients for this empirical risk Recent ideas from graph sampling theory to (i) define an empirical risk for Blei, Peter Orbanz Download PDF Abstract: Empirical risk minimization is the main tool for prediction problems, but itsĮxtension to relational data remains unsolved. Our results show that both theoretically and empirically, objective perturbation is superior to the previous state-of-the-art, output perturbation, in managing the inherent tradeoff between privacy and learning performance.Authors: Victor Veitch, Morgane Austern, Wenda Zhou, David M. We obtain encouraging results from evaluating their performance on real demographic and benchmark data sets. We apply these results to produce privacy-preserving analogues of regularized logistic regression and support vector machines. We further present a privacy-preserving technique for tuning the parameters in general machine learning algorithms, thereby providing end-to-end privacy guarantees for the training process. If the loss and regularizer satisfy certain convexity and differentiability criteria, we prove theoretical results showing that our algorithms preserve privacy, and provide generalization bounds for linear and nonlinear kernels. This method entails perturbing the objective function before optimizing over classifiers. Then we propose a new method, objective perturbation, for privacy-preserving machine learning algorithm design. First we apply the output perturbation ideas of Dwork et al. These algorithms are private under the ε-differential privacy definition due to Dwork et al. We provide general techniques to produce privacy-preserving approximations of classifiers learned via (regularized) empirical risk minimization (ERM). Privacy-preserving machine learning algorithms are crucial for the increasingly common setting in which personal data, such as medical or financial records, are analyzed. Kamalika Chaudhuri, Claire Monteleoni, Anand D. Differentially Private Empirical Risk Minimization
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |