Simulation results of non-energy based neural networks for scheduling job shops

Most neural network models for solving job-shop scheduling (JSS) problems are energy-based and usually take a long time to converge to solutions. In the authors, prior work, they proposed a new neural model called job-shop scheduling neural networks (JSSNNs), which need no special convergence procedure to be performed and can find optimal or near-optimal solutions of large JSS problems at a much faster speed. Furthermore, this new model takes significantly fewer numbers of neurons and interconnections, which are 5p and 9p respectively, where p is the number of operations. In the worst case, the model takes mn(m+3) neurons and mn(n²+n+5) interconnections to solve an n-job m-machine problem. However, the number of operations that can be handled is mn. Thus, it is very feasible to implement the model in hardware for solving large problems. In this paper, the authors present simulation results of this new type of neural network and a simple method that obtains a lower bound of a schedule for appraising its quality