This page presents selected visualizations resulting from research conduced with my postdocs and students. Note: Part of my research is in pattern recognition and image analysis; see for instance the images resulting from our research on OCT Imaging.
The image below presents the outcome of detecting blood vessels in a fundus image (conventional method for assessment of the back of the eye) using a detector evolved by means of genetic programming. See this paper for details.
An optimized two-dimensional toroidal embedding of all 132 unique semantics obtained from 1024 programs of depth 3 composed of instructions +,−, , /, v, 1. Each tile plots program semantics in the interval [−1, 1]. Shading between tiles is the darker the greater the semantic distance. See (krawiec:2011:EuroGP) and (DBLP:conf/gecco/Krawiec11) for details.
Improving search efficiency using related problems
Evolution of small-board Go player strategy (a hybrid of coevolutionary learning and temporal difference learning). See (2010CECKrawiecSzubert) for details.
See (2009LichockiKrawiecEvoGames) for details.
(click to pause, double-click to resume)
If all you can see is 'This plugin is not supported', your browser must be missing a plugin to play Quicktime movies. You can alternatively download the movie using this link.