Timetable scheduling using neural networks with parallel implementation on transputers

A scheduling neural network based on the interactive activation and competition model, integrating into the conventional sequential scheduling framework to perform timetable scheduling is proposed. Parallel implementation of the model of a transputer system is realized by task decomposition so that independent groups of classes are scheduled by independent transputers and conflicts are resolved via message passing on priorities of classes. Compared to the Hopfield network-based optimization approach, the size of the scheduling network considered grows at a slower rate (linear) with the problem size and is flexible for encoding more realistic constraints