Realization of Autonomous Soft Computing System Using Computational Intelligence Methods
Rohit Kant Srivastava, Jwalant Baria · 2012
Computational Intelligence or Soft Computing methods refer to three concepts viz. Artificial Neural Network, Fuzzy Systems and Evolutionary Algorithm (EA) or Genetic Algorithm (GA). The Computational Intelligence methodologies facilitate a better data acquisition from our everyday environment, and thus help in extending the capabilities of conventional computer systems. The fuzzy controller is realized with the application of fuzzy logic concepts to an information processing system. The controller is based on linguistic rules, so that human experts can insert knowledge into the system. The behaviour of neural networks is adapted by input data supplied by the environment. These networks learn by adjusting the strength of connections (weights) between the neurons. The entire system turns out to be complex and is optimized by application of GA. 1. COMPUTATIONAL INTELLIGENCE The three Soft Computing methods operating in synchronization with each other for real time learning: fuzzy systems, artificial neural networks and genetic algorithms are referred as Computational Intelligence (CI) methods. Unlike classical reasoning and modeling approaches, these methods provide a set of more flexible computing technique for solving real-world problems involving incomplete, uncertain and imprecise input- output information. There are two well established approaches to handle this imperfect information. One is the probabilistic reasoning approach, based on probability theory. The other is possibilistic reasoning approach, based on the semantics of fuzzy sets and many-valued logics. Both disciplines differ in terminology and apparatus. The theory of probability serves to determine a degree of uncertainty, in which any of the presumed effects occurs. In the contrary fuzzy logic helps to determine which one of many effects defined in different way have occurred. The aim of the implementation of both methods is to obtain better control and prediction of highly non-linear systems behaviour.