July 8, 2026
Dimitris Kelesis, researcher in the DICE group of the Institute of Informatics and Telecommunications, has been awarded the Best PhD Thesis in Artificial Intelligence in Greece for 2026 by the Hellenic Artificial Intelligence Society (EETN).
His thesis, titled “Efficient Approaches to Deal with Oversmoothing in Deep Graph Neural Networks”, was selected among seven high-quality submissions following a competitive evaluation process by the EETN scientific committee.
The work addresses a key open problem in Graph Neural Networks, namely oversmoothing, proposing a novel metric and providing a rigorous analysis of the behaviour and characteristics of deep graph models. The thesis combines strong mathematical foundations with extensive experimental validation, offering a comprehensive framework for understanding and improving graph-based learning models.
Its results are expected to have significant impact, contributing to the development of more reliable, efficient and interpretable models for graph-structured data, with applications across a wide range of domains.
The thesis was jointly supervised by the School of Electrical and Computer Engineering of the National Technical University of Athens (NTUA), and the Institute of Informatics and Telecommunications, under the co-supervision of Prof. Dimitris Fotakis and Dr. Giorgos Paliouras, who serves as Acting Director of the Institute.
This distinction further strengthens a consistent track record of excellence within the Institute’s research community. In 2024, Periklis Mantenoglou received the Runner-up Honourable Mention, while in 2023 Vangelis Micheloudakis was awarded the Best PhD Thesis in AI in Greece. Notably, all three theses were developed under joint supervision arrangements between the Institute and the Universities, a reflection of the Institute’s enduring academic partnerships.
This continuity highlights the sustained effort towards high-quality research and reflects both the achievements of the Institute’s students and the strength of its research environment.
The full PhD thesis is available in the National Documentation Centre repository.