Title :
An LMI-neural network based solution to the load balancing problem for heterogeneous local clusters
Author :
Silva, João M M ; Kaszkurewicz, Eugenius
Author_Institution :
Brazilian Naval Res. Inst., Rio de Janeiro
Abstract :
A solution for the load balancing problem in local clusters of heterogeneous processors is proposed within the setting of delayed artificial neural networks, optimal control and linear matrix inequalities (LMI) theory. Based on a mathematical model that includes delays and processors with different processing velocities, this model is transformed into a special case of delayed cellular neural networks model. A systematic method of controller synthesis is derived, based on two coupled linear matrix inequalities - one guaranteeing global convergence and the other guaranteeing performance in the linear region of operation. Simulations and computational experiments show the efficiency of this approach, reducing load balancing time.
Keywords :
cellular neural nets; control system synthesis; linear matrix inequalities; neurocontrollers; optimal control; pattern clustering; resource allocation; LMI-neural network based solution; artificial neural networks; controller synthesis; delayed cellular neural networks; heterogeneous local clusters; heterogeneous processors; linear matrix inequalities; load balancing problem; load balancing time; mathematical model; optimal control; Artificial neural networks; Cellular neural networks; Computational modeling; Control system synthesis; Convergence; Linear matrix inequalities; Load management; Mathematical model; Network synthesis; Optimal control;
Conference_Titel :
Neural Networks, 2008. IJCNN 2008. (IEEE World Congress on Computational Intelligence). IEEE International Joint Conference on
Conference_Location :
Hong Kong
Print_ISBN :
978-1-4244-1820-6
Electronic_ISBN :
1098-7576
DOI :
10.1109/IJCNN.2008.4634116
