Get our free extension to see links to code for papers anywhere online!

 Add to Chrome

 Add to Firefox

CatalyzeX Code Finder - Browser extension linking code for ML papers across the web! | Product Hunt Embed

Decentralized Differentially Private Without-Replacement Stochastic Gradient Descent

Sep 12, 2018
Richeng Jin, Xiaofan He, Huaiyu Dai

While machine learning has achieved remarkable results in a wide variety of domains, the training of models often requires large datasets that may need to be collected from different individuals. As sensitive information may be contained in the individual's dataset, sharing training data may lead to severe privacy concerns. Therefore, there is a compelling need to develop privacy-aware machine learning methods, for which one effective approach is to leverage the generic framework of differential privacy. Considering that stochastic gradient descent (SGD) is one of the mostly adopted methods for large-scale machine learning problems, two decentralized differentially private SGD algorithms are proposed in this work. Particularly, we focus on SGD without replacement due to its favorable structure for practical implementation. In addition, both privacy and convergence analysis are provided for the proposed algorithms. Finally, extensive experiments are performed to verify the theoretical results and demonstrate the effectiveness of the proposed algorithms.

* The Abstract is revised 

Share this with someone who'll enjoy it:

   Access Paper Source

Share this with someone who'll enjoy it: