Homepage of the Research Group

Intelligent Data Analysis

with Machine Learning, Neural Networks and Fuzzy Logic

at the

Institute of Computer Design and Fault Tolerance,

Department of Computer Science

University of Karlsruhe, Germany

NEW: Lecture Intelligente Datenanalyse

Members of our group:

• Dr. Fridtjof Feldbusch
• Dr. Klaus-Peter Huber
• Dr. Michael R. Berthold

Research Interests:

Computers have proven to be very helpful to solve a lot of tasks involving products of our daily life, in factories controlling production, or even organizing management activities. To do so, Computers use special programs that explicitly control the order of steps to solve the tasks they are used for. The drawback of this approach is that these programs are not flexible enough to work when the environment changes for which they were developed. This inflexibility is often not acceptable and leads to a rising demand for easy and automatic adaptivity for tasks in changing environments or even to changing tasks. We think that present adaption techniques are not sufficiently evolved to meet expectations people have.
That is why we are working on the following subjects:

• Constructive algorithms for Local Basis Function Networks (M. Berthold):
  The Dynamic Decay Adjustment Algorithm (DDA) allows to train Radial Basis Functions fast and efficient. Performance has been demonstrated to be comparable to state of the art classifiers.
• Rule Extraction from Neural Networks (M. Berthold, K.-P. Huber):
  A new approach called Rectangular Basis Function Networks has been proposed. These types of networks allow to extract the acquired knowledge in the form of if-then-rules.
• Intelligent Data Analysis (K.-P. Huber, M. Berthold):
  Analysis of huge datasets using techniques of the area of Machine Learning and combinations of Neural Networks with Fuzzy Logic offer insights into the underlying structure.
• Neuro Fuzzy (F Feldbusch):
  Combinations of Fuzzy Logic with conventional Neural Networks allow a prestructuring of Neural Networks using Fuzzy rules.

Points of Interest:

newsgroups:

COMP.AI.NEURAL-NETS
COMP.AI.FUZZY

Online Information:

UIUC AI Information List
Wavelets...
Neuroscience Web Search

WWW-sites of other groups:

University of Amsterdam Robotics and Neurocomputing
Neuro Fuzzy Uni Braunschweig
UKA: The NN learning algorithm benchmarking page
CMU, CNBC

WWW-sites of journals, societies, companies:

IEEE Neural Networks Council (NNC)
International Neural Network Society
GI Fachgruppe 1.1.3 "Maschinelles Lernen"

Machine Learning Online
The NIPS-Homepage
NIPS*95 Online papers
Neural Processing Letters
Neural Computation

NeuroDimension, Inc. - Application Corner

F. Feldbusch, 29.6.1998