From Krzysztof Krawiec

Site: CIG

Computational Intelligence Group

CIG is an informal group of my closest collaborators within the Laboratory of Intelligent Decision Support Systems, which currently includes postdocs (Wojciech Jaśkowski, Bartosz Wieloch, Marcin Szubert, Tomasz Pawlak, Paweł Liskowski), and Phd students (Iwo Błądek, Jakub Bednarek and Jakub Kwiatkowski).

We are primarily interested in machine learning, pattern recognition, software synthesis, and have experience in using the following CI paradigms:

• conventional machine learning,
• evolutionary computation and genetic programming,
• neural networks, including deep learning,
• reinforcement learning, especially when hybridized with coevolutionary algorithms.

Our notable achievements include:

• State-of-the-art algorithms for program synthesis based on the paradigm of genetic programming: Our paper ''Behavioral Programming: A Broader and More Detailed Take on Semantic GP", co-authored with Una-May O'Reilly of CSAIL, MIT, won the Best Paper Award in the Genetic Programming track of the GECCO'14 conference.
• Deep neural network approach that provides best-to-date segmentation of retinal blood vessel networks in ophthalmological imaging (fundus imaging), outperforming over 70 other methods designed to date. More details in our paper Segmenting Retinal Blood Vessels with Deep Neural Networks in IEEE Transactions on Medical Imaging.
• Top-ranking Othello strategies trained in a knowledge-free manner (i.e., without explicit use of domain knowledge): our position evaluation function (epTDLmpx_12x6), ranked first in the Othello Position Evaluation Function League, organized by Simon Lucas and his team at University of Essex, UK.

CIG publications are available on my publication page and on the pages of individual members.