Gaussian-Based Runtime Detection of Out-of-distribution Inputs for Neural Networks

HASHEMI, Vahid; Jan KŘETÍNSKÝ; Stefanie MOHR a Emmanouil SEFERIS. Gaussian-Based Runtime Detection of Out-of-distribution Inputs for Neural Networks. In Runtime Verification - 21st International Conference, RV 2021, Virtual Event, October 11-14, 2021, Proceedings. Springer, 2021, s. 254-264. ISBN 9783030884932. Dostupné z: https://doi.org/10.1007/978-3-030-88494-9_14.

Základní údaje

Originální název

Gaussian-Based Runtime Detection of Out-of-distribution Inputs for Neural Networks

Autoři

HASHEMI, Vahid; Jan KŘETÍNSKÝ; Stefanie MOHR a Emmanouil SEFERIS

Vydání

Runtime Verification - 21st International Conference, RV 2021, Virtual Event, October 11-14, 2021, Proceedings, od s. 254-264, 11 s. 2021

Nakladatel

Springer

---

Další údaje

Typ výsledku

Stať ve sborníku

Označené pro přenos do RIV

Ne

Organizační jednotka

Fakulta informatiky

ISBN

9783030884932

ISSN

DOI

https://doi.org/10.1007/978-3-030-88494-9_14

---

Anotace

V originále

In this short paper, we introduce a simple approach for runtime monitoring of deep neural networks and show how to use it for out-of-distribution detection. The approach is based on inferring Gaussian models of some of the neurons and layers. Despite its simplicity, it performs better than recently introduced approaches based on interval abstractions which are traditionally used in verification.