MeStE: Multi-dimensional Statistical Classification

Project Topic

Analysis, visualization and application of classification techniques to multi-dimensional data

Project Description

Analysis and comparison of statistical classification procedures with a focus on the following aspects:

• extraction of relevant input factors (dimension reduction)
• computation and visualization of effects
• robustness and stability of methods
• quantification of differences between the procedures

Project Members

Project Chair

• Prof. Dr. Ralf Korn
• Prof. Dr. Jürgen Franke
• Prof. Dr. Dieter Prätzel-Wolters

Participating Research Groups

• Financial Mathematics and Stochastic Controlling Group (Department of Mathematics)
• Applied Mathematical Statistics Group (Department of Mathematics)
• Prediction and Decision Support Group (Cluster University of Kaiserslautern - Saarland University)

Scientific Personnel

• Dipl.Phys. Andreas Jablonski (Sect. Models and Algorithms in Image Processing, Fraunhofer Institute for Industrial Mathematics (ITWM))
• Dr. Hagen Knaf (Sect. System Analysis, Prognosis and Control (formerly Sect. Adaptive Systems), Fraunhofer Institute for Industrial Mathematics (ITWM))
• PD Dr. Marlene Müller (Sect. Financial Mathematics, Fraunhofer Institute for Industrial Mathematics (ITWM))
• Dipl.-Inform. Markus Rauhut (Sect. Models and Algorithms in Image Processing, Fraunhofer Institute for Industrial Mathematics (ITWM))

External Cooperation

• Prof. Dr. Michael G. Schimek (Med.Universität Graz, Österreich)

Project Events and Achievements

• Project start date: January 1st, 2006
• Project end date: December 31st, 2007

Project Workshop

• Workshop "Klassifikationsverfahren für die Anwendung in Medizin, Wirtschaft und Technik" [Flyer], September 20, 2007 at Fraunhofer ITWM
  
  Programm (9:30-18:00, siehe [Flyer])
  • Begrüßung (Ralf Korn)
  • Einführung (Hagen Knaf, Marlene Müller, Markus Rauhut, ITWM)
  • Top-k-Ranglisten: Aggregation und Inferenz (Michael G. Schimek, Uni Graz & Uni Brünn)
  • Nichtparametrische Komponenten in Discrete-Choice-Modellen (Marlene Müller, ITWM)
  • Large margin classification (Alexander Zien, MPI Tübingen)
  • ROC-basierte Diskriminanzanalyse (Hagen Knaf, ITWM)
  • Prediction of response to combination antiretroviral therapy for HIV infection (Andre Altmann, MPI Saarbrücken)
  • Classification of spectral data using random forests (Bjoern Menze, Uni Heidelberg)
  • Clustering in der Bildverarbeitung (Andreas Jablonski & Oliver Wirjadi, ITWM)

Presentations

• Marlene Müller, Michael G. Schimek: Analysis of Highdimensional Data by Semiparametric (Generalized) Regression Models, invited talk, Interface 2006, Pasadena/California (USA), May 2006. [Slides]
• Marlene Müller: KernGPLM - A Package for Kernel-Based Fitting of Generalized Partial Linear and Additive Models, useR 2006, Vienna (Austria), June 2006. [Abstract] [Slides]
• Hagen Knaf: Kern-Fisher-Diskriminanzanalyse aus analytischer Sicht, internal ITWM seminar, June 2006 [Slides]

Project Publications

Kernel Fisher discriminant functions - a concise and rigorous introduction

Hagen Knaf. Technical Report, Berichte des ITWM, Volume 117, ITWM, Kaiserslautern, January, 2007

Soft Operators Decision Trees - Uncertainty and Stability related Issues

E. Barrena. Technische Universität Kaiserslautern, Dissertation, 2007

Automatische Konfiguration von Bildverarbeitungssystemen - Konzept und Anwendungsbeispiele

S. Peters, A. König. In: ITG-Fachgruppensitzung "Mikroelektronik Neuronaler Netze", ISE, Kaiserslautern. November, 2006

Comparison and Visualization of Statistical Classifikation Methods for Credit Defaults

Tanja Dankert. Masters Thesis, September, 2006

Statistical Aspects of Setting up a Credit Rating System

Beatriz Clavero Rasero. September, 2006

Automated feature selection for the classification of meningioma cell nuclei

Oliver Wirjadi, Thomas Breuel, Wolfgang Feiden, Yoo-Jin Kim. In: Springer ed., Bildverarbeitung für die Medizin, Informatik aktuell. P. 76-80, 2006