Stavroulakis P. (ed.) Neuro-Fuzzy and Fuzzy-Neural Applications in Telecommunications

Springer, 2004. — 352 p.Neurofuzzy and fuzzyneural techniques as tools of studying and analyzing complex problems are relatively new even though neural networks and fuzzy logic systems have been applied as computational intelligence structural elements for the last 40 years.Computational intelligence as an independent scientific field has grown over the years because of the development of these structural elements.
Neural networks have been revived since 1982 after the seminal work of J.J. Hopfield and fuzzy sets have found a variety of applications since the publication of the work of Lotfi Zadeh back in 1965. Artificial neural networks (ANN) have a large number of highly interconnected processing elements that usually operate in parallel and are configured in regular architectures. The collective behavior of an ANN, like a human brain, demonstrates the ability to learn, recall, and generalize from training patterns or data. The performance of neural networks depends on the computational function of the neurons in the network, the structure and topology of the network, and the learning rule or the update rule of the connecting weights. This concept of trainable neural networks further strengthens the idea of utilizing the learning ability of neural networks to learn the fuzzy control rules, the membership functions and other parameters of a fuzzy logic control or decision systems, as we will explain later on, and this becomes the advantage of using a neural based fuzzy logic system in our analysis.
On the other hand, fuzzy systems are structured numerical estimators. They start from highly formalized insights about the psychology of categorization and the structure of categories in the real world and they articulate fuzzy IF-THEN rules as a kind of expert knowledge. As a general principle, fuzzy logic is based on the incompatibility principle which suggests that complexity and ambiguity are correlated. As we learn more and more about a system, its complexity decreases and our understanding increases. The major tasks encountered in using fuzzy systems involve determination of fuzzy logic rules and the membership functions. These fuzzy systems which are based on these two basic elements are also called fuzzy inference systems. The fuzzy logic rules and membership functions of the fuzzy system can be used to find and interpret the structure and the weights of neural networks. Fuzzy neural networks have higher training speed and are more and more robust than conventional neural systems.
In general, integrating fuzzy systems with ANNs and ANNs with fuzzy systems, we maximize the learning and adaptive capabilities of the combined system which are not available in neither of the systems from which the integrated system came from. Which of the two tools (neurofuzzy of fuzzyneural) will be used in a particular situation depends on the particular application. If for example some expert information is available at the outset and is easy to develop some rules, we can use a neurofuzzy system for which the neural network has been used to further refine the fuzzy rules.Integration of Neural and Fuzzy
Neuro-Fuzzy Applications in Speech Coding and Recognition
Image/Video Compression Using Neuro-Fuzzy Techniques
A Neuro-Fuzzy System for Source Location and Tracking in Wireless Communications
Fuzzy-Neural Applications in Handoff
An Application of Neuro-Fuzzy Systems for Access Control in Asynchronous Transfer Mode Networks
A: Overview of Neural Networks
B: Overview of Fuzzy Logic Systems
C: Examples of Fuzzy-Neural and Neuro-Fuzzy Integration