Summary

Adversarial learning aims to identify weaknesses in machine learning models. The goal is to identify potential problems that cannot be found using traditional evaluation using test sets. It has been used successfully in a wide range of applications, typically focused on a specific model or domain. In image classification, methods have been developed to fool models that recognize traffic signs by rather simple modifications of pictures. Another direction of adversarial learning aims to identify examples that could break or improve the training of the model if that example would be added to the training. In this project, we will extend recently introduced methods that identify adversarial examples and prevent a miss-classification of these examples. These methods are inspired by the concept of applicability domain used in cheminformatics.

Duration and Type

  • As a student project preferable 1 semester, but can be also done in 2 semesters
  • The summer scholarship is over 12 weeks between S2 2020 and S1 2021
  • Honours project, other postgraduate project (MSc, MProfStuds, …)

Requirements

  • Basic maths, CS, and machine learning skills
  • Programming in Java or Python

Supervisor and contact

  • Joerg Wicker
  • Send CV and transcript by mail to Joerg Wicker
  • Applications for the summer scholarships via Faculty of Science