Saved in:
Bibliographic Details
Main Authors: McMahan, H. Brendan, Xu, Zheng, Zhang, Yanxiang
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2408.08868
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916766701060096
author McMahan, H. Brendan
Xu, Zheng
Zhang, Yanxiang
author_facet McMahan, H. Brendan
Xu, Zheng
Zhang, Yanxiang
contents The state-of-the-art for training on-device language models for mobile keyboard applications combines federated learning (FL) with differential privacy (DP) via the DP-Follow-the-Regularized-Leader (DP-FTRL) algorithm. Two variants of DP-FTRL are used in practice, tree aggregation and matrix factorization. However, tree aggregation suffers from significantly suboptimal privacy/utility tradeoffs, while matrix mechanisms require expensive optimization parameterized by hard-to-estimate-in-advance constants, and high runtime memory costs. This paper extends the recently introduced Buffered Linear Toeplitz (BLT) mechanism to multi-participation scenarios. Our BLT-DP-FTRL maintains the ease-of-use advantages of tree aggregation, while essentially matching matrix factorization in terms of utility and privacy. We evaluate BLT-DP-FTRL on the StackOverflow dataset, serving as a re-producible simulation benchmark, and across four on-device language model tasks in a production FL system. Our empirical results highlight the advantages of the BLT mechanism and elevate the practicality and effectiveness of DP in real-world scenarios.
format Preprint
id arxiv_https___arxiv_org_abs_2408_08868
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle A Hassle-free Algorithm for Private Learning in Practice: Don't Use Tree Aggregation, Use BLTs
McMahan, H. Brendan
Xu, Zheng
Zhang, Yanxiang
Machine Learning
The state-of-the-art for training on-device language models for mobile keyboard applications combines federated learning (FL) with differential privacy (DP) via the DP-Follow-the-Regularized-Leader (DP-FTRL) algorithm. Two variants of DP-FTRL are used in practice, tree aggregation and matrix factorization. However, tree aggregation suffers from significantly suboptimal privacy/utility tradeoffs, while matrix mechanisms require expensive optimization parameterized by hard-to-estimate-in-advance constants, and high runtime memory costs. This paper extends the recently introduced Buffered Linear Toeplitz (BLT) mechanism to multi-participation scenarios. Our BLT-DP-FTRL maintains the ease-of-use advantages of tree aggregation, while essentially matching matrix factorization in terms of utility and privacy. We evaluate BLT-DP-FTRL on the StackOverflow dataset, serving as a re-producible simulation benchmark, and across four on-device language model tasks in a production FL system. Our empirical results highlight the advantages of the BLT mechanism and elevate the practicality and effectiveness of DP in real-world scenarios.
title A Hassle-free Algorithm for Private Learning in Practice: Don't Use Tree Aggregation, Use BLTs
topic Machine Learning
url https://arxiv.org/abs/2408.08868