Hybrid Systems of Computational Intelligence Evolved from Self- Learning Spiking Neural Network

Computational intelligence paradigm covers several approaches for technical problems solving in an intelligence manner, such as artificial neural networks, fuzzy logic systems, evolutionary computation, etc. Each approach provides engineers and researchers with the smart and powerful tools to handle various real-life concerns. Even more powerful tools were designed at the joint of different computational intelligence approaches. Neuro-fuzzy systems, for example, are well-known and advanced intelligent tool that combines capabilities of neural networks and fuzzy systems together in a synergetic way. Among them, one of the prominent hybrid systems type is self-learning fuzzy spiking neural networks. They were evolved from fuzzy logic systems and selflearning spiking neural networks, and revealed considerable computational capabilities. There were proposed several architectures of self-learning fuzzy spiking neural networks, each handling a particular kind of data processing tasks (processing fuzzy data, fuzzy probabilistic and possibilistic clustering, batch and adaptive methods, new clusters detection, irregular form clusters detection, etc). In this paper, known architectures of self-learning hybrid systems based on spiking neural network are reviewed, compared, and summarized. A generalized architecture and learning algorithm for self-learning fuzzy spiking neural networks are proposed.