A Neural Network Realization of File Transfer Scheduling

Neural network models have been successfully applied to solve a variety of problems requiring associative recall, constraint satisfaction, and optimization. This paper presents a new scheduling approach based upon a deterministic modified Hopfield model to solve "File-Transfer" scheduling, an NP-Complete constraint satisfaction problem. The proposed model is mapped onto a 2-dimensional neural network architecture for the transfer scheduling of files between various nodes of a network, by which the overall transfer times is to be minimized. Neural Network-based Scheduling is achieved by formulating the scheduling problem in terms of energy function, and by using the "Motion Equation" corresponding to the variation of energy levels. The main contribution of this work is an efficient and fast parallel algorithm under time and resource constraints, appropriate for implementation on the parallel machines. However, neurons' motion equation is the core of this guided movement mechanism which searches the scheduling space in parallel, and guarantees that the state of system mostly converges to the optimum state. Yet another important contribution of this work is the new strategy of constraints and variables containments by which the performance and efficacy of the system was considerably improved.