by Georg Krempl, Daniel Kottke Mining data with minimal annotation costs requires efficient active approaches, that ideally select the optimal candidate for labelling under a user-specified classification performance measure. Common generic approaches, that are usable with any classifier and any performance measure, are either slow like error reduction, or heuristics like uncertainty sampling. In contrast, […]
by Georg Krempl, Daniel Kottke, Myra Spiliopoulou. Active Mining of Big Data requires fast approaches that ideally select for a user-specified performance measure and arbitrary classifier the optimal instance for improving the classification performance. Existing generic approaches are either slow, like error reduction, or heuristics, like uncertainty sampling. We propose a novel, fast yet versatile […]
by Daniel Kottke, Gino Gulamhussene. Performing statistically significant analysis with medical images requires data from many test persons. Using medical imaging, this is only applicable by reducing quality. Furthermore, research on vertebrae deformation should not rely on top-down (model-based) methods. Our proposed algorithm automatically detects the central spinal curve with 3D data-driven methods on multi-sequence […]
I am tutor at the summerschool 2013 at University Magdeburg for a lecture on reinforcement learning. The student’s task is to implement an agent’s artificial intelligence for a framework I developed. The agent has to find the green goal as fast as possible.
The importance of video annotation is constantly rising, especially in areas of multimedia retrieval, semantic search and even in newest research in neuroscience. The list of pixel-accurate, hierarchical and automatic object detection tools is still empty, due to the excessive complexity. The approach is to roll out the automation to a problem specific detection algorithm. […]