Self-certifying classification by linearized deep assignment
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-...
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Dokumenttyp: | Article (Journal) Kapitel/Artikel |
| Sprache: | Englisch |
| Veröffentlicht: |
18 Feb 2022
|
| Ausgabe: | Version v2 |
| In: |
Arxiv
Year: 2022, Pages: 1-19 |
| DOI: | 10.48550/arXiv.2201.11162 |
| Online-Zugang: | Verlag, kostenfrei, Volltext: https://doi.org/10.48550/arXiv.2201.11162 Verlag, kostenfrei, Volltext: http://arxiv.org/abs/2201.11162 
|
| Verfasserangaben: | Bastian Boll, Alexander Zeilmann, Stefania Petra, Christoph Schnörr |
| Zusammenfassung: | 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. |
|---|---|
| Beschreibung: | Online veröffentlicht am 26. Januar 2022 Gesehen am 10.01.2024 |
| Beschreibung: | Online Resource |
| DOI: | 10.48550/arXiv.2201.11162 |