Straggler-Resilient Differentially-Private Decentralized Learning
This work addresses privacy and efficiency challenges in decentralized machine learning for distributed systems, but it is incremental as it builds on existing differential privacy frameworks.
The paper tackles the straggler problem in decentralized learning by extending differential privacy amplification to include training latency, deriving convergence and privacy results for skipping and baseline schemes, and identifying a trade-off validated on logistic regression and image classification datasets with concrete metrics like accuracy and latency.
We consider the straggler problem in decentralized learning over a logical ring while preserving user data privacy. Especially, we extend the recently proposed framework of differential privacy (DP) amplification by decentralization by Cyffers and Bellet to include overall training latency--comprising both computation and communication latency. Analytical results on both the convergence speed and the DP level are derived for both a skipping scheme (which ignores the stragglers after a timeout) and a baseline scheme that waits for each node to finish before the training continues. A trade-off between overall training latency, accuracy, and privacy, parameterized by the timeout of the skipping scheme, is identified and empirically validated for logistic regression on a real-world dataset and for image classification using the MNIST and CIFAR-10 datasets.