Dr. Dominik Belter, Poznań University of Technology
“Gait control of the six-legged robot on a rough terrain using computational intelligence learning and optimization methods”
Supervisor: Prof. Piotr Skrzypczyński

The dissertation addresses the problem of planning the motion of a six-legged robot walking in rough and previously unknown environments. The robot does not have a map of the terrain in advance, thus it has to use a dedicated measurement system. The analytical methods which are used to control the robot are supported by computational intelligence learning and optimization methods.

Dr. Piotr Andruszkiewicz, Warsaw University of Technology
“Privacy Preserving Classification and Association Rules Mining over Centralised Data”
Supervisor: Prof. Marzena Kryszkiewicz

The thesis is devoted to privacy preserving classification and association rules mining over centralised data distorted with randomisation-based methods which modify individual values at random to provide an expected level of privacy. The new privacy preserving solutions proposed in the thesis reduce time complexity of association rules mining, enable miners to use ordered attributes and achieve higher accuracy of classification.

Dr. Szymon Chojnacki, Institute of Computer Science, Polish Academy of Sciences
“Analiza technicznych własności systemów rekomendujących za pomocą grafów losowych”
Supervisor: Prof. Mieczysław Kłopotek

I evaluate technical properties of recommender systems. Such systems are used in e-commerce. Their task is to propose interesting items for a particular user. One of largest problems in using recommender systems is long time of preparing a recommendation. I evaluate this problem and propose new techniques of assessing the complexity of recommender systems. I also propose new methods of implementing the algorithms.

Dr. Przemysław Więch, Warsaw University of Technology
“Distributed Default Reasoning in the Semantic Web”
Supervisor: Prof. Henryk Rybiński

The dissertation presents a novel approach to distributed reasoning in a multi-agent system, in which the agents' knowledge bases are expressed in the language of description logic with defaults. The thesis proposes a formalism of default transformations, which can be used for passing messages containing transformed defaults between agents. This solution allows for efficient message passing and avoids reaching conclusions based on assumptions that are accepted too early.