| Summary: | 碩士 === 國立勤益科技大學 === 電子工程系 === 98 === In recent years, the concept of scheduling has been widely applied to various applications. How to get a proper scheduling; in terms of decision-making managers has become more and more important. Most of the scheduling application problems are confirmed to be the NP-complete problems. Therefore, a variety of research methods are introduced in solving these scheduling problems, such as fuzzy logic, neural network (NN), genetic algorithm (GA), ant colony optimization (ACO) algorithm and particle swarm optimization (PSO) algorithm. Moreover, many integrations of above stated schemes are also proposed to improve the efficiency. Among them, fuzzy Hopfield neural network which integrated fuzzy c-means into Hopfield neural network to become a famous clustering method. However, most applications using FHNN are limited in solving two dimensional problems.
In this work, a 2-D fuzzy Hopfield neural network (FHNN) is expanded into a three dimensional (3-D) network for solving the scheduling problems of multiprocessor scheduling problems. Restated, a processor is regarded as a cluster in multivariable scheduling problem. Moreover, three defuzzification strategies are applied and simulated; they are competition rule, roulette wheel selection rule (random-proportional rule) and pseudo-random-proportional rule.
This investigation employs the suggested approach to resolve multiprocessor scheduling problems with time constrains (execution time and deadline) and no migration allowed. Simulation results illustrate that the proposed 3-D fuzzy Hopfield neural network involving proposed defuzzification strategies provides an appropriate approach for solving this class of scheduling problems effectively.
|
|---|
