基于深度学习的资源投入问题算法

Deep learning algorithm for resource investment problem

针对资源投入调度问题,提出了基于实时调度状态的调度优先级规则智能决策机制,构造了嵌合人工神经网络的双层迭代循环搜索算法。算法上层为启发式资源搜索框架,下层为基于实时调度状态的调度优先级规则智能决策算法。下层算法通过双隐层BP神经网络离线学习,获得调度状态与调度优先级规则的映射关系,并在实时调度过程中的每一阶段,根据当前调度数据,智能决策调度优先级规则,并指导作业调度进行。最后,通过标准算例库PSPLIB进行对比实验,验证了所设计算法的有效性。

An intelligent decision-making scheme of scheduling priority rules based on real-time scheduling state was presented for resource and job scheduling of resource investment problem,and a double-layer iterative cyclic search algorithm based on artificial neural network was proposed.The upper stage of the algorithm was a heuristic resource search framework,and the lower stage was an intelligent decision-making algorithm of scheduling priority rules based on real-time scheduling status.The lower stage of algorithm obtained the mapping relationship between scheduling status and scheduling priority rules through off-line learning of double hidden layer BP neural network.The scheduling priority rules were decided intelligently at each stage of real-time scheduling process,which guided job scheduling according to current scheduling data.The effectiveness of the designed algorithm was verified by comparison with other literature algorithm through experiments with PSPLIB.