Private Adaptive Optimization with Side information

Tian Li, Manzil Zaheer, Sashank Reddi, Virginia Smith
Proceedings of the 39th International Conference on Machine Learning, PMLR 162:13086-13105, 2022.

Abstract

Adaptive optimization methods have become the default solvers for many machine learning tasks. Unfortunately, the benefits of adaptivity may degrade when training with differential privacy, as the noise added to ensure privacy reduces the effectiveness of the adaptive preconditioner. To this end, we propose AdaDPS, a general framework that uses non-sensitive side information to precondition the gradients, allowing the effective use of adaptive methods in private settings. We formally show AdaDPS reduces the amount of noise needed to achieve similar privacy guarantees, thereby improving optimization performance. Empirically, we leverage simple and readily available side information to explore the performance of AdaDPS in practice, comparing to strong baselines in both centralized and federated settings. Our results show that AdaDPS improves accuracy by 7.7% (absolute) on average—yielding state-of-the-art privacy-utility trade-offs on large-scale text and image benchmarks.

Cite this Paper

BibTeX
@InProceedings{pmlr-v162-li22x, title = {Private Adaptive Optimization with Side information}, author = {Li, Tian and Zaheer, Manzil and Reddi, Sashank and Smith, Virginia}, booktitle = {Proceedings of the 39th International Conference on Machine Learning}, pages = {13086--13105}, year = {2022}, editor = {Chaudhuri, Kamalika and Jegelka, Stefanie and Song, Le and Szepesvari, Csaba and Niu, Gang and Sabato, Sivan}, volume = {162}, series = {Proceedings of Machine Learning Research}, month = {17--23 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v162/li22x/li22x.pdf}, url = {https://proceedings.mlr.press/v162/li22x.html}, abstract = {Adaptive optimization methods have become the default solvers for many machine learning tasks. Unfortunately, the benefits of adaptivity may degrade when training with differential privacy, as the noise added to ensure privacy reduces the effectiveness of the adaptive preconditioner. To this end, we propose AdaDPS, a general framework that uses non-sensitive side information to precondition the gradients, allowing the effective use of adaptive methods in private settings. We formally show AdaDPS reduces the amount of noise needed to achieve similar privacy guarantees, thereby improving optimization performance. Empirically, we leverage simple and readily available side information to explore the performance of AdaDPS in practice, comparing to strong baselines in both centralized and federated settings. Our results show that AdaDPS improves accuracy by 7.7% (absolute) on average—yielding state-of-the-art privacy-utility trade-offs on large-scale text and image benchmarks.} }
Endnote
%0 Conference Paper %T Private Adaptive Optimization with Side information %A Tian Li %A Manzil Zaheer %A Sashank Reddi %A Virginia Smith %B Proceedings of the 39th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2022 %E Kamalika Chaudhuri %E Stefanie Jegelka %E Le Song %E Csaba Szepesvari %E Gang Niu %E Sivan Sabato %F pmlr-v162-li22x %I PMLR %P 13086--13105 %U https://proceedings.mlr.press/v162/li22x.html %V 162 %X Adaptive optimization methods have become the default solvers for many machine learning tasks. Unfortunately, the benefits of adaptivity may degrade when training with differential privacy, as the noise added to ensure privacy reduces the effectiveness of the adaptive preconditioner. To this end, we propose AdaDPS, a general framework that uses non-sensitive side information to precondition the gradients, allowing the effective use of adaptive methods in private settings. We formally show AdaDPS reduces the amount of noise needed to achieve similar privacy guarantees, thereby improving optimization performance. Empirically, we leverage simple and readily available side information to explore the performance of AdaDPS in practice, comparing to strong baselines in both centralized and federated settings. Our results show that AdaDPS improves accuracy by 7.7% (absolute) on average—yielding state-of-the-art privacy-utility trade-offs on large-scale text and image benchmarks.
APA
Li, T., Zaheer, M., Reddi, S. & Smith, V.. (2022). Private Adaptive Optimization with Side information. Proceedings of the 39th International Conference on Machine Learning, in Proceedings of Machine Learning Research 162:13086-13105 Available from https://proceedings.mlr.press/v162/li22x.html.

Related Material