New optimizer DP-IVON-Gradsq enhances differential privacy in Bayesian deep learning

Researchers have developed DP-IVON-Gradsq, a new optimizer designed to enhance differential privacy in Bayesian deep learning. This method aims to mitigate the interference between privacy noise and the stochasticity inherent in Bayesian posterior sampling. By using a noise-corrected squared-gradient estimator, DP-IVON-Gradsq maintains computational efficiency similar to Adam while improving privacy guarantees. Evaluations on the CIFAR-10 dataset indicate that DP-IVON-Gradsq performs competitively with DP-SGD and DP-Adam under less stringent privacy constraints. AI

IMPACT Introduces a novel optimizer that could improve privacy-preserving training of Bayesian deep learning models.

RANK_REASON The cluster contains a research paper detailing a new method for differential privacy in Bayesian deep learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Adam
• arXiv
• Bayesian deep learning
• CIFAR-10
• differential privacy
• DP-Adam
• DP-IVON-Gradsq
• DP SGD
• Improved Variational Online Newton

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →