Improved Convergence of Differential Private SGD with Gradient Clipping

Authors: Huang Fang, Xiaoyun Li, Chenglin Fan, Ping Li

ICLR 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on standard benchmark datasets are conducted to support our analysis.
Researcher Affiliation Industry Cognitive Computing Lab Baidu Research No.10 Xibeiwang East Road, Beijing 100193, China 10900 NE 8th St. Bellevue, Washington 98004, USA {fangazq877,lixiaoyun996,fanchenglin,pingli98}@gmail.com
Pseudocode Yes Algorithm 1 Differential-private SGD with gradient clipping (DP-SGD-GC)
Open Source Code No The paper does not provide a specific link or explicit statement about releasing its own source code for the methodology described.
Open Datasets Yes We conduct experiments on two standard image classification benchmark datasets: MNIST (Le Cun, 1998) and CIFAR10 (Krizhevsky & Hinton, 2009).
Dataset Splits No The paper mentions training and testing accuracy but does not specify a validation dataset split.
Hardware Specification Yes All experiments are conducted on a server with 4 CPUs and one NVIDIA Tesla P100 GPU.
Software Dependencies No The paper mentions 'Paddle Paddle' and 'Opacus package' but does not specify version numbers for these or other software dependencies.
Experiment Setup Yes For all experiments, we set the batch size B = 128, the noise level σ = 1.0 and the confidence level δ = 10^-5. For MNIST, we try learning rate in {2e-3, 5e-3, 1e-2} for each experiment and report the best result. For CIFAR10 we fix the learning rate to be 0.1.