Table of Contents

Marcin Dziubiński

2011 Polish Artificial Intelligence Society Award for the Best Ph. D. Dissertation in Artificial Intelligence

Dr. Marcin Dziubiński, Warsaw University
Complexity issues in multimodal logics for multiagent systems
supervisor: dr hab. Barbara Dunin-Kęplicz

The dissertation studies computational complexity of satisfiability problem for TeamLog – a formalism for cooperating agents designed by Barbara Dunin-Kęplicz and Rineke Verbrugge. It shows that TeamLog complexity is exponential. Consequently language restrictions are sought. Novel and universal modal context restriction is proposed. It obtains sufficiently expressive TeamLog fragments and enables usage of other restrictions or heuristic approaches.



Wojciech Jaśkowski

Honourable mention in the competition for the 2011 Polish Artificial Intelligence Society Award for the Best Ph. D. Dissertation in Artificial Intelligence

Dr. Wojciech Jaśkowski, Poznań University of Technology
Algorithms for Test-based Problems
supervisor: dr hab. inż. Krzysztof Krawiec

The dissertation is devoted to a subclass of optimization problems, characterized by the fact that potential solutions are evaluated on a (usually large) set of elementary tests. The work proposes new computationally effective co-evolutionary algorithms to solve such problems. Obtained results have applications in learning game strategies, machine learning, and combinatorial optimization.



Joanna Czajkowska

Honourable mention for the Best Applied Thesis in the competition for the 2011 Polish Artificial Intelligence Society Award for the Best Ph. D. Dissertation in Artificial Intelligence

Dr. Joanna Czajkowska, Politechnika Śląska
Parameterization and 3D Segmentation of Bone Tumours in Magnetic Resonance Images
supervisor: prof. dr hab. inż. Ewa Piętka

Thesis is available on request from the author.

The dissertation presents a three dimensional method of bone tumour segmentation in magnetic resonance images. The segmentation method presented is applicable to different types of bone tumours and lesions. It is also insensitive to their location in human body. The proposed segmentation method is based on Kernelized Fuzzy C-Means clustering, Gaussian Mixture Model and Relative Fuzzy Connectedness analysis.