Scientia Publishing House – Sapientia Hungarian University of Transylvania

Neurális hálózatok és fuzzy logika [Neural Networks and Fuzzy Logic]

The volume entitled Neural Networks and Fuzzy Logic is divided into three parts: neural networks, fuzzy logic, and hardware implementations. Throughout 9 chapters deal with the basics of neural networks, and in two chapters the fuzzy logic and fuzzy inference systems are discussed. In the first part of the book within the part of neural networks, the basics of neural network topologies, neural network training solution, single-layer perceptron structures, and training are presented. In the fourth chapter, the gradient-based optimization algorithm used for neural network parameter update, the cost functions used for neural network training, solutions for data normalization, and regularization solutions for neural network size optimization are discussed. Chapter five describes the structure of the multilayer perceptron and its teaching with the backpropagation algorithm. In Chapter 6, the RBF neural network structure, teaching, and application areas are detailed.

In Chapter 7, competitive neural nets are presented. In the framework of competitive neural nets, the Kohonen network topology and training as well as the application of the Kohonen Network for the travelling salesman problem are detailed. In the next chapter, associative and auto-associative nets are being discussed. The structure and teaching of the CMAC-type neuronal network are explained in Chapter 9. In the 10th and 11th chapters, the basic knowledge of fuzzy logic, membership functions, fuzzy set operations, fuzzy relation with fuzzy set, and Mamdani- and Takagi-Sugeno-type fuzzy inference systems are reviewed. Chapter 11 provides insights to the application of the Mamdani-type fuzzy inference system. In the third part, we discuss the hardware-based implementation of neural networks and fuzzy inference systems. The constraints applied to the hardware implementation of neural networks, such as applied arithmetic, precision, and possible parallelization solutions, are detailed.

In the final chapters, the hardware implementation of an RBF neuron network and a Takagi-Sugeno fuzzy inference system are presented. The results related to the hardware implementation were achieved during the scientific research I have carried out. In the book, I also tried to apply the experience gained during the teaching in the field of artificial intelligence at Sapientia University.