• Media type: E-Article
  • Title: Self‐certifying classification by linearized deep assignment
  • Contributor: Boll, Bastian [Author]; Zeilmann, Alexander [Author]; Petra, Stefania [Author]; Schnörr, Christoph [Author]
  • Published: Augsburg University Publication Server (OPUS), 2023
  • Language: English
  • DOI: https://doi.org/10.1002/pamm.202200169
  • Origination:
  • Footnote: Diese Datenquelle enthält auch Bestandsnachweise, die nicht zu einem Volltext führen.
  • Description: We propose a novel class of deep stochastic predictors for classifying metric data on graphs within the PAC-Bayes risk certification paradigm. Classifiers are realized as linearly parametrized deep assignment flows with random initial conditions. Building on the recent PAC-Bayes literature and data-dependent priors, this approach enables (i) to use risk bounds as training objectives for learning posterior distributions on the hypothesis space and (ii) to compute tight out-of-sample risk certificates of randomized classifiers more efficiently than related work. Comparison with empirical test set errors illustrates the performance and practicality of this self-certifying classification method.
  • Access State: Open Access