Thesis supervisor: Márk Jelasity
Location of studies (in Hungarian): University of Szeged Abbreviation of location of studies: SZTE
Description of the research topic:
Modern artificial intelligence systems are composed of components that are optimizied or created using some form of machine learning. Machine learning typically means approximating a function based on a finite set of examples. In the past few years, it has become clear that these automatically learned components have strange properties. In general, it is not clear how they arrive at certain decisions, and they can be mislead using artificially generated adversarial inputs. For example, adding specially designed invisible noise to images can change the predicted label arbitrarily. The research problem consists of developing techniques and approaches that increase the robustness of different machine learning models (and the systems based on these models) against different adversarial attacks, and also to develop tools and techniques to increase the interpretability of these models in a well-defined manner. Also, the study of the relationship of interpretability and robustness is especially interesting, as it is likely that these two properties are interrelated.
Required language skills: English Number of students who can be accepted: 3
Deadline for application: 2022-03-15
2024. IV. 17. ODT ülés Az ODT következő ülésére 2024. június 14-én, pénteken 10.00 órakor kerül sor a Semmelweis Egyetem Szenátusi termében (Bp. Üllői út 26. I. emelet).
Login
