Fuzzy Data Analysis

Uncertainty and Vagueness in Knowledge Based Systems

Neural Networks and Fuzzy Systems Neuro-Fuzzy Systems

In our group we work on theoretical foundations of neuro-fuzzy systems, and practical realizations of these concepts. Neural networks can learn from data, but cannot be interpreted - they are black boxes to the user. Fuzzy Systems consist of interpretable linguistic rules, but they cannot learn. We use learning algorithms from the domain of neural networks to create fuzzy systems from data. The learning algorithms can learn both fuzzy sets, and fuzzy rules, and can also use prior knowledge.

What are neuro-fuzzy systems? Read about our view of these combinations of fuzzy systems with neural network methods.

Based on our research activities we have developed several neuro-fuzzy models:

• The NEFCON model (NEuro Fuzzy CONtrol) supports the development of fuzzy controllers by reinforcement learning. The learning process uses a fuzzy error measure that can be defined by the user. NEFCON is available in three different implementations (for Unix, Windows, and MATLAB/SIMULINK).
• NEFCLASS is short for NEuro Fuzzy CLASSification. It is a neuro-fuzzy model for data analysis. From a set of training data it can learn classifications rules, and suitable fuzzy sets. NEFCLASS is available for MS-DOS and Unix systems.
• NEFPROX is a model for NEuro Fuzzy function apPROXimation. it is similar to the two models above, but it is more general. You can use it to learn a fuzzy system from training data that is able to approximate the function given by the data samples. NEFPROX will be availiable for Unix systems in the near future.

If you want to know more about our work please have a look at our list of publications, or choose a paper from our list of papers (with abstracts) and download it.

For further information please contact Detlef Nauck, Andreas Nürnberger, or Rudolf Kruse.