Self‐Certifying Classification by Linearized Deep Assignment

Abstract: 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.

Object Details

• Location: Deutsche Nationalbibliothek Frankfurt am Main
• Extent: Online-Ressource
• Language: Englisch

References and Relationships

• Bibliographic citation: Self‐Certifying Classification by Linearized Deep Assignment ; volume:23 ; number:1 ; year:2023 ; extent:6 ; Proceedings in applied mathematics and mechanics ; 23, Heft 1 (2023) (gesamt 6)

Contributors, Places and Time

• Creator: Boll, Bastian ; Zeilmann, Alexander ; Petra, Stefania ; Schnörr, Christoph

Further information

• DOI: 10.1002/pamm.202200169
• URN: urn:nbn:de:101:1-2023060115141258953453
• Rights: Open Access; Der Zugriff auf das Objekt ist unbeschränkt möglich.
• Last update: 14.08.2024, 6:52 PM CEST