PrivSGP-VR: Differentially Private Variance-Reduced Stochastic Gradient Push with Tight Utility Bounds

Authors: Zehan Zhu, Yan Huang, Xin Wang, Jinming Xu

IJCAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments corroborate our theoretical findings, especially in terms of the maximized utility with optimized K, in fully decentralized settings.
Researcher Affiliation Academia 1Zhejiang University, Hangzhou, China 2Qilu University of Technology, Jinan, China
Pseudocode Yes Algorithm 1 Priv SGP-VR
Open Source Code No The paper does not explicitly state that its source code is open-sourced or provide a direct link to a code repository. It only refers to a 'full version' on arXiv.
Open Datasets Yes We consider two non-convex learning tasks (i.e., deep CNN Res Net-18 [He et al., 2016] training on Cifar-10 dataset [Krizhevsky, 2009] and shallow 2-layer neural network training on Mnist dataset [Deng, 2012]), in fully decentralized setting.
Dataset Splits No The paper states 'For all experiments, we split shuffled datasets evenly to n nodes.' but does not provide specific training/validation/test split percentages, absolute sample counts, or explicit methodology for these splits.
Hardware Specification Yes All experiments are deployed in a high performance computer with Intel Xeon E5-2680 v4 CPU @ 2.40GHz and 8 Nvidia RTX 3090 GPUs
Software Dependencies No The paper mentions 'Py Torch' and 'torch.distributed' but does not specify their version numbers.
Experiment Setup Yes All configurations utilize the same DP Gaussian noise variance σ2 i = 0.03 for each node i. For each node i, we set the privacy budget to ϵi = 3 and δi = 10 5. ... we apply DP Gaussian noise with an identical variance of σ2 i = 0.03 for both Priv SGP-VR and Priv SGP. Moreover, both algorithms were executed for a fixed number of 3000 iterations.