Research Interests
Machine Learning
- Inductive Learning
- Version Space Learning
- Supervised Learning of concept descriptions for High-Level Scene Interpretation
- Learning logic-based concept descriptions
- Learning probability-based concept descriptions
- Representing spatial arrangements
- Concept differentiation
Scene Interpretation
- Learning a comprehensive knowledge base of ontological concepts
- Employing learnt concepts in an interpretation system
- Automatic evaluation of learnt concepts
- Confidence-based evaluation of learnt concepts
Other
- Cognitive Science
- Learning Theory
- Chess Programming
Projects
eTraining for Interpreting Images of Man-Made Scenes (link)
The aim of this project is to advance the state of the art of cognitive systems by developing a methodology for autonomous and continuous learning. The project will concentrate on structural learning, where relations between components and compositional hierarchies play a central role in object categorization. Such learning is particularly relevant for the interpretation of man-made objects, hence the project will use the recognition of buildings in outdoor scenes as its exemplary application domain.
The ontology presented above is learnt automatically using a Version Space Framework. All concepts are desribed through a rich description language employing logic-based and probabilistic attribute types. Spatial Relations are represented qualitative and quantitative.
Involvement in the project: Apr 2006 - Apr 2009
Publications
2007
Johannes Hartz; Bernd Neumann: Version Space Learning of Ontological Structures for High-level Scene Interpretation, Technical Report TR FBI-B-277/07, Department of Informatics, University of Hamburg, Sep 2007.
Johannes Hartz; Bernd Neumann: Learning a Knowledge Base of Ontological Concepts for High-Level Scene Interpretation, IEEE Proc. International Conference on Machine Learning and Applications 2007, Cincinnati (Ohio, USA), Dec 2007.
