Neural Network Techniques: A Tutorial on Interconnection, Learning and Stability

The subject of neural networks is approached from the point of view architecture and theory. Starting from the human brain as inspiration, the neural network is constructed step by step from artificial neuron towards a complete trained network able to perform its task. Every aspect such as topology, interconnection, learning and stability is considered. The most significant issues of four frequently used neural networks are discussed. These are: the common known multilayer feedforward neural network, the Hopfield neural network, the Kohonen neural network, important in robot control and the radial basis function network. The latter has recently become very important in automatic control, especially in identification of unknown processes.