Timezone: »
Existing approaches for training neural networks with user-level differential privacy (e.g., DP Federated Averaging) in federated learning (FL) settings involve bounding the contribution of each user's model update by {\em clipping} it to some constant value. However there is no good {\em a priori} setting of the clipping norm across tasks and learning settings: the update norm distribution depends on the model architecture and loss, the amount of data on each device, the client learning rate, and possibly various other parameters. We propose a method wherein instead of a fixed clipping norm, one clips to a value at a specified quantile of the update norm distribution, where the value at the quantile is itself estimated online, with differential privacy. The method tracks the quantile closely, uses a negligible amount of privacy budget, is compatible with other federated learning technologies such as compression and secure aggregation, and has a straightforward joint DP analysis with DP-FedAvg. Experiments demonstrate that adaptive clipping to the median update norm works well across a range of federated learning tasks, eliminating the need to tune any clipping hyperparameter.
Author Information
Galen Andrew (Google)
Om Thakkar (Google)
Brendan McMahan (Google Research)
Swaroop Ramaswamy (Google)
More from the Same Authors
-
2020 : Understanding Unintended Memorization in Federated Learning »
Om Thakkar -
2021 Poster: Revealing and Protecting Labels in Distributed Training »
Trung Dang · Om Thakkar · Swaroop Ramaswamy · Rajiv Mathews · Peter Chin · Françoise Beaufays -
2020 : Contributed Talk #7: Training Production Language Models without Memorizing User Data »
Swaroop Ramaswamy · Om Thakkar -
2020 Tutorial: (Track1) Federated Learning and Analytics: Industry Meets Academia Q&A »
Peter Kairouz · Brendan McMahan · Virginia Smith -
2020 Poster: Privacy Amplification via Random Check-Ins »
Borja Balle · Peter Kairouz · Brendan McMahan · Om Thakkar · Abhradeep Guha Thakurta -
2020 Tutorial: (Track1) Federated Learning and Analytics: Industry Meets Academia »
Brendan McMahan · Virginia Smith · Peter Kairouz -
2019 : Privacy for Federated Learning, and Federated Learning for Privacy »
Brendan McMahan -
2019 Workshop: Workshop on Federated Learning for Data Privacy and Confidentiality »
Lixin Fan · Jakub Konečný · Yang Liu · Brendan McMahan · Virginia Smith · Han Yu -
2018 : Poster Session »
Phillipp Schoppmann · Patrick Yu · Valerie Chen · Travis Dick · Marc Joye · Ningshan Zhang · Frederik Harder · Olli Saarikivi · Théo Ryffel · Yunhui Long · Théo JOURDAN · Di Wang · Antonio Marcedone · Negev Shekel Nosatzki · Yatharth A Dubey · Antti Koskela · Peter Bloem · Aleksandra Korolova · Martin Bertran · Hao Chen · Galen Andrew · Natalia Martinez · Janardhan Kulkarni · Jonathan Passerat-Palmbach · Guillermo Sapiro · Amrita Roy Chowdhury -
2018 : Brendan McMahan »
Brendan McMahan -
2018 Poster: Graph Oracle Models, Lower Bounds, and Gaps for Parallel Stochastic Optimization »
Blake Woodworth · Jialei Wang · Adam Smith · Brendan McMahan · Nati Srebro -
2018 Spotlight: Graph Oracle Models, Lower Bounds, and Gaps for Parallel Stochastic Optimization »
Blake Woodworth · Jialei Wang · Adam Smith · Brendan McMahan · Nati Srebro -
2018 Poster: Model-Agnostic Private Learning »
Raef Bassily · Abhradeep Guha Thakurta · Om Thakkar -
2018 Poster: cpSGD: Communication-efficient and differentially-private distributed SGD »
Naman Agarwal · Ananda Theertha Suresh · Felix Xinnan Yu · Sanjiv Kumar · Brendan McMahan -
2018 Spotlight: cpSGD: Communication-efficient and differentially-private distributed SGD »
Naman Agarwal · Ananda Theertha Suresh · Felix Xinnan Yu · Sanjiv Kumar · Brendan McMahan -
2018 Oral: Model-Agnostic Private Learning »
Raef Bassily · Abhradeep Guha Thakurta · Om Thakkar -
2014 Poster: Delay-Tolerant Algorithms for Asynchronous Distributed Online Learning »
Brendan McMahan · Matthew Streeter -
2013 Poster: Minimax Optimal Algorithms for Unconstrained Linear Optimization »
Brendan McMahan · Jacob D Abernethy -
2013 Poster: Estimation, Optimization, and Parallelism when Data is Sparse »
John Duchi · Michael Jordan · Brendan McMahan -
2012 Poster: No-Regret Algorithms for Unconstrained Online Convex Optimization »
Matthew Streeter · Brendan McMahan