Dagstuhl-Seminar 9702
Theory and Practice of Machine Learning
( 06. Jan – 10. Jan, 1997 )
Permalink
Organisatoren
- H.U. Simon (Dortmund)
- R. Sutton (Stowe Research)
- T. Dietterich (Corvallis)
- W. Maass (Graz)
Kontakt
This Seminar brought together researchers from various branches of computer science whose work involves learning by machines. It provided an opportunity for theoreticans to measure their models and results against demands and data that arise in applied machine learning, and it provided an opportunity for researchers in applied machine learning to initiate a more systematic and rigorous theoretical investigation of new learning approaches that have emerged from experimental work on application problems.
One new learning paradigm that had received considerable attention at this Seminar is the Reinforcement Learning approach. This approach takes into account that in many practical machine learning scenarios no source of classified training examples is available to the learner. Another novel type of learning problems arises in the context of complex probabilistic models, which have turned out to be particularly useful for representing knowledge in a variety of real-world application domains. Apart from these two new directions the abstracts of this report demonstrate the vitality, sophistication and success of current work in theoretical and applied machine learning.
Most of the abstracts contain pointers to homepages where details of the reported work can be found. Pointers are also given to tutorials by Mike Jordan, Hans Simon und Rich Sutton, which provide introductions to some of the research areas discussed at this Seminar.
We would like to thank Lucas Palette for collecting these abstracts, and for providing pointers to the homepages of participants.
- H.U. Simon (Dortmund)
- R. Sutton (Stowe Research)
- T. Dietterich (Corvallis)
- W. Maass (Graz)