PAC-Bayesian Generalization Bounds

Ask about this topic

Overview

PAC-Bayesian generalization bounds control the gap between training and test risk by treating a learned predictor as a distribution over models and charging for how far that distribution moves from a prior. Common approaches use KL-divergence bounds for Gibbs predictors, with tighter variants based on data-dependent priors, compression, parameter sensitivity, flatness, or the geometry of equivalent predictors. The framework is being extended to settings such as deep networks, graph and quantum models, control, and dependent time-series data, where bounds can guide training or provide deployment certificates. A central tension is making guarantees both non-vacuous and computationally useful without relying on restrictive assumptions about the data, prior, or predictor.

Latest papers

CardsList

35 papers5 papers in the last four weeks, down 17% on the four weeks before. 0.0% of all new papers.

  1. Distributionally Robust PAC-Bayesian Control

    Apr 12, 2026Domagoj Herceg, Duarte AntunesDistribution Shift RobustnessPAC-Bayesian Generalization Bounds

    The authors are with the Department of Mechanical Engineering, Eindhoven university of Technology, Eindhoven, The Netherlands.

  2. Topology-Aware PAC-Bayesian Generalization Analysis for Graph Neural Networks

    Apr 12, 2026Xinping YiNeural Network GeneralizationPAC-Bayesian Generalization Bounds

    School of Information Science and Engineering, Southeast University, Nanjing, China

  3. A PAC-Bayesian approach to generalization for quantum models

    Mar 24, 2026Pablo Rodriguez-Grasa, Matthias C. Caro, Jens Eisert +3Generalization BoundsPAC-Bayesian Generalization Bounds

    Department of Physical Chemistry, University of the Basque Country UPV/EHU, 48080 Bilbao, Spain · EHU Quantum Center, University of the Basque Country UPV/EHU, 48080 Bilbao, Spain · TECNALIA, Basque Research and Technology Alliance (BRTA), 48160 Derio, Spain +4

  4. Overfitting and Generalizing with (PAC) Bayesian Prediction in Noisy Binary Classification

    Mar 23, 2026Xiaohan Zhu, Mesrob I. Ohannessian, Nathan SrebroPAC-Bayesian TheoryBinary Classification

    The University of Chicago · The University of Illinois at Chicago · Toyota Technological Institute at Chicago

  5. How Learning Dynamics Drive Adversarially Robust Generalization?

    Oct 10, 2024Yuelin Xu, Xiao ZhangAdversarial TrainingPAC-Bayesian Generalization Bounds

    CISPA Helmholtz Center for Information Security, Saarbrücken, Germany