{"slug": "from-privacy-to-generalization-linear-max-information-bounds-for-dp-sgd", "title": "From Privacy to Generalization: Linear Max-Information Bounds for DP-SGD", "summary": "Researchers have proven a finite-sample bound on the approximate max-information of differentially private stochastic gradient descent (DP-SGD) that scales at most linearly with dataset size, matching classic results for ε-differentially private algorithms. The finding establishes a direct link between privacy and generalization in deep learning, yielding explicit PAC-Bayes generalization bounds where the prior can be learned by DP-SGD and a complexity term controlled by optimization hyperparameters. This resolves a long-standing open problem in machine learning theory by providing the first generalization guarantees for DP-SGD-trained models with fully explicit complexity terms.", "body_md": "arXiv:2605.26222v1 Announce Type: new\nAbstract: Understanding the relationship between generalization and privacy remains a central challenge in modern machine learning theory, particularly for deep networks trained by variants of differentially private stochastic gradient descent (DP-SGD). In this work we make progress on this persistent open problem by proving a finite-sample bound on the approximate max-information of DP-SGD that exhibits scaling properties comparable with (Dwork et al, 2015)'s classic result for $\\epsilon$-differentially private algorithms, namely at most linear in the dataset size. From our result we obtain a general-purpose PAC-Bayes generalization bound in which the necessary prior distribution can be learned by DP-SGD, as well as a generalization bound for DP-SGD-trained models themselves, with a complexity term that is fully explicit and controlled by the optimization hyperparameters.", "url": "https://wpnews.pro/news/from-privacy-to-generalization-linear-max-information-bounds-for-dp-sgd", "canonical_source": "https://arxiv.org/abs/2605.26222", "published_at": "2026-05-27 04:00:00+00:00", "updated_at": "2026-05-27 04:29:48.824897+00:00", "lang": "en", "topics": ["machine-learning", "neural-networks", "ai-research", "ai-safety"], "entities": ["Dwork et al"], "alternates": {"html": "https://wpnews.pro/news/from-privacy-to-generalization-linear-max-information-bounds-for-dp-sgd", "markdown": "https://wpnews.pro/news/from-privacy-to-generalization-linear-max-information-bounds-for-dp-sgd.md", "text": "https://wpnews.pro/news/from-privacy-to-generalization-linear-max-information-bounds-for-dp-sgd.txt", "jsonld": "https://wpnews.pro/news/from-privacy-to-generalization-linear-max-information-bounds-for-dp-sgd.jsonld"}}