Neural networks for SISD real time control

Due to complexity of neural networks the following problems appear: (1) extensive requirements on computational devices where the network is trained or evaluated, (2) slow network evaluation/training response. The problem is more evident when using networks on devices where conventional algorithms run without problems with respect to the time response. In general this problem has to be solved using a more powerful computational device or parallel systems. This paper deals with a neural network optimized for running on a single (SISD) end-device to obtain real time response in real time control systems and in this paper special effort is made to introducing the associated problems and designed solutions.