Detection of misclassified and adversarial examples in Deep Learning

Marco Blanco, Jorge (2019). Detection of misclassified and adversarial examples in Deep Learning. Thesis (Master thesis), E.T.S. de Ingenieros Informáticos (UPM).

Description

Title: Detection of misclassified and adversarial examples in Deep Learning
Author/s:
  • Marco Blanco, Jorge
Contributor/s:
  • Baumela Molina, Luis
Item Type: Thesis (Master thesis)
Masters title: Inteligencia Artificial
Date: 2019
Subjects:
Faculty: E.T.S. de Ingenieros Informáticos (UPM)
Department: Inteligencia Artificial
Creative Commons Licenses: Recognition - No derivative works - Non commercial

Full text

[img]
Preview
PDF - Requires a PDF viewer, such as GSview, Xpdf or Adobe Acrobat Reader
Download (715kB) | Preview

Abstract

Great developments have been carried out thanks to deep learning in recent years. However, for a massive adoption in real-world applications, deep learning models need to be trustworthy. In this work we identify scenarios where such models behave in an unexpected fashion, implement a method based on the entropy for assessing their reliability and build a deep learning-based system able to detect its own wrong predictions. Additionally, by using the mentioned method, we achieve a relevant improvement of accuracy with an ensemble of learners.

More information

Item ID: 56004
DC Identifier: http://oa.upm.es/56004/
OAI Identifier: oai:oa.upm.es:56004
Deposited by: Biblioteca Facultad de Informatica
Deposited on: 09 Aug 2019 06:38
Last Modified: 09 Aug 2019 06:38
  • Logo InvestigaM (UPM)
  • Logo GEOUP4
  • Logo Open Access
  • Open Access
  • Logo Sherpa/Romeo
    Check whether the anglo-saxon journal in which you have published an article allows you to also publish it under open access.
  • Logo Dulcinea
    Check whether the spanish journal in which you have published an article allows you to also publish it under open access.
  • Logo de Recolecta
  • Logo del Observatorio I+D+i UPM
  • Logo de OpenCourseWare UPM