Transformation knowledge in pattern analysis with kernel methods [Elektronische Ressource] : distance and integration kernels / von Bernard Haasdonk

Dissertationzur Erlangung des Doktorgradesder Fakulta¨t fu¨r Angewandte Wissenschaftenan der Albert-Ludwigs-Universitat Freiburg im Breisgau¨Transformation Knowledge inPattern Analysis with KernelMethods–Distance and Integration KernelsvonDipl.-Math. Bernard Haasdonk24. Mai 2005iiDekan: Prof. Dr. Jan G. KorvinkPrufungskommission: Prof. Dr. Wolfram Burgard (Vorsitz)¨Prof. Dr. Luc De Raedt (Beisitz)Prof. Dr. Hans Burkhardt (Gutachter)Prof. Dr. Bernhard Scholkopf (Gutachter)¨Datum der Disputation: 18. November 2005AcknowledgementFirstly, IwanttothankmysupervisorProf.Dr.-Ing.HansBurkhardtforgivingmethepossibilityandwidesupportfortheresearchwhichhasledtothisthesis. Inparticular,the excellent technical environment, the availability of various interesting applicationfields and the scientific freedom have combined to be an excellent basis for indepen-dent research. The generous support of research travel enabled me to establish manyimportant and fruitful contacts. Similarly, I am deeply grateful to Prof. Dr. BernhardSch¨olkopf who was a constant source of motivation through his own related work andvarious guiding hints, many of which find themselves realized in the present thesis. Iam very glad that he agreed to act as the second referee. In particular, I am verythankful for being given the opportunity to visit his group for a talk, several weeksof research and the machine learning summer school MLSS 2003.